I’ve been on both sides of that conversation. When I started in recruitment two years ago, I genuinely wasn’t sure. Now I’d say I can pick it out most of the time without any tool at all. The shift hasn’t been in whether AI writing exists; it’s that the question has moved on. It’s now less “can this be detected?” and more “what are people using to detect it, and is Turnitin even relevant here?”

For most hiring teams, Turnitin isn’t relevant. It’s a product built for universities, sold on institutional licences, priced and designed around academic submission workflows. If you work in HR and you’re looking at the Turnitin AI detector alternatives that actually come up in recruitment conversations, you’re looking at a different set of tools entirely.

The short answer: most hiring teams aren’t using Turnitin. The tools that come up in recruitment are Walter Writes AI Detector, Proofademic, GPTZero, and Originality.ai. For HR specifically, you want something fast and affordable that doesn’t require you to set up an institutional account. Turnitin isn’t built for that.

Turnitin AI Detector Alternatives: 6 Tools Recruiters Use in 2026

This is based on what shows up in HR discussions online and what I’ve tested myself. It’s not exhaustive. What works for a high-volume agency is going to be completely different from what makes sense for a small internal team.

1. Walter Writes

I use Walter Writes more than anything else on this list, and the main reason isn’t what most people would guess. The use case for me isn’t primarily checking applicant text. It’s making sure my own drafts don’t flag when I’m putting together outreach messages or writing up job posts, because I use AI assistance for those and I want to check my own output before it goes out.

The AI humanizer and AI detector are in the same editor. That’s what actually makes it useful day to day. You’re not jumping between tabs or logging into separate tools; you run the draft through the humanizer, check the score, and that’s it.

For incoming text, it works fine too. Paste a cover letter in and you get a score that’s been benchmarked against Turnitin, GPTZero, Originality.ai, and Copyleaks. One check gives you a sense of where it lands across the main detectors.

2. Proofademic

Proofademic comes up most in conversations about education-adjacent hiring, which makes sense. It was built for academic contexts, and its main differentiator is that you get sentence-level analysis rather than one overall score. So instead of “73% AI”, you can see exactly which sentences flagged and what the probability was for each one. For any role where you’re actually assessing writing quality, that’s far more useful.

The other feature that matters is the Paraphrase Shield. It catches text that’s been paraphrased after AI generation. Most tools don’t do this, and at this point, running a draft through QuillBot before submitting is basically the standard workaround. Proofademic picks up on it.

3. aichecker.tech

Free, no account required. Good for a quick one-off check when you don’t have budget for a subscription and you just want a signal. The analysis is pretty basic compared to the paid tools, but it does the job for occasional use.

4. aidetector.ac

Same category as aichecker.tech: free, no registration, gets you a score quickly. People in my network who use it tend to do so because there’s genuinely no setup involved. Useful if you just need something for the odd check without committing to anything.

5. GPTZero

GPTZero was one of the first standalone AI detectors and it’s still widely used. You get probability scores at both document and sentence level, and the accuracy is decent on unedited AI text. The integrations and the UI were clearly built for educators rather than recruiters, which it was. It works for hiring purposes, just not optimised for it.

6. Originality.ai

The main thing that sets Originality.ai apart is that it combines AI detection with plagiarism checking. That’s useful if you’re looking at content portfolios for writing-heavy roles. Credit-based pricing rather than a subscription makes it better for occasional use than as a daily screening tool.

AI Detector for HR: What the Screening Process Looks Like in Practice

Most recruitment teams I know aren’t running a formalised AI detection process at all. What’s happening is more informal than that.

After reading enough applications in a short period of time, you start to develop a feel for it. The writing structure starts looking the same across applications. The vocabulary is technically correct but it has this impersonal quality, like something a company PR team would write rather than a person. The way enthusiasm is expressed reads like it’s describing the candidate’s excitement in the third person. You don’t need software for that. You need to read a lot of applications in a short space of time.

Where the tools come in is in specific use cases: graduate scheme applications where writing quality is part of what’s being assessed, comms or marketing roles where writing is the actual job, or in organisations that have made a deliberate policy decision to screen for AI use. For those situations, Proofademic tends to come up most in my corner of the internet, followed by GPTZero and Originality.ai.

The sentence-level breakdown you get from Proofademic is what makes it more useful for hiring specifically. You want to know which section of the document flagged, not just that the overall score was high. I wrote more about the recruiter side of this in my earlier piece on Turnitin AI detection in cover letters and what recruiters are noticing in 2026.

One thing I’d add: if your organisation hasn’t landed on a clear policy about AI screening, I’d be careful about building detection into your workflow on your own initiative. False positives on human-written text happen more than people expect. If you’re making decisions about candidates based on a score, you’re taking on real risk.

Turnitin vs Proofademic: Which One Works Better for Hiring Teams

The fundamental problem with Turnitin in an HR context is that it wasn’t designed for HR. The licensing model is built around institutions, the integrations are built around academic submission workflows, the support and documentation assumes you’re a university. You can technically use it outside academia, but it’s awkward.

A few differences that matter practically:

Proofademic has a free 1,000-word trial, no credit card needed, and the paid tier starts at $99 per year. Turnitin requires institutional licensing. For a recruitment team or an independent recruiter, the starting point alone makes Turnitin impractical.

On detection quality: Turnitin gives you an overall AI score. Proofademic’s sentence-level detection breaks it down by sentence, with reasoning for why each one flagged. When you’re making a hiring decision, that distinction matters. You can see if the flagged content is in the most important part of the application or in a generic opener that half of applicants use.

Proofademic also has the Paraphrase Shield, which catches rewritten AI text. Turnitin doesn’t have an equivalent. Given how common QuillBot and similar tools are as workarounds, that’s a meaningful gap.

For a recruitment team doing occasional checks, Proofademic is the more practical choice. It’s built for a similar detection purpose but without the institutional overhead.

Best Plagiarism Checker Alternative in 2026: What’s Worth Using

Worth being clear on what these tools actually do, because “plagiarism checker” gets used to mean a few different things.

A traditional plagiarism checker runs text against a database of existing content to find matches. An AI detector looks for statistical patterns in the writing that indicate machine generation. These are different problems with different tools. Some products do both (Originality.ai being the main example), but most do one or the other.

If you’re trying to figure out what’s worth using in 2026, it comes down to being clear about what you’re trying to find.

Walter Writes’ free AI detector gives you AI-likelihood scores benchmarked against Turnitin, GPTZero, and the other main detectors. You don’t need an account to use the free version, and it’s in the same editor as the humanizer, which is useful if you’re checking your own work as well as incoming content.

For screening applicant submissions specifically, Proofademic’s sentence-level output and paraphrase detection make it better suited to that use case than most alternatives.

Tools like Grammarly’s plagiarism checker or Copyscape are doing something different. They’re looking for text that’s been copied from somewhere else. If the problem you’re trying to solve is AI-generated writing, those tools won’t surface it. I went into the content creation side in more detail in AI Tools for Recruitment Writing in 2026: What We Actually Use.

Does Turnitin Detect Walter Writes AI Output?

This comes up in the HR groups I’m in fairly regularly, and it’s nearly always candidates asking rather than recruiters.

The answer genuinely depends on what’s happened to the text. Raw AI output, the text that’s come straight out of a model with no editing, still carries the statistical patterns that detectors are trained on. The two main ones are perplexity (predictable word choices, low variation in vocabulary) and burstiness (uniform sentence lengths throughout). Unedited ChatGPT output scores high on both.

Text that’s been through a proper structural rewrite is a different story. Walter Writes’ humanizer works at the structural level rather than just swapping synonyms, which is what most basic “paraphrase” tools do. Their before/after examples show Turnitin going from 95% AI-detected to 100% Human. GPTZero and Originality.ai scores move in the same direction.

For candidates reading this: if you’ve run your text through a structural humanizer and done a reasonable editing pass yourself, the scanner probably isn’t your main problem. The bigger risk is a recruiter who’s read three hundred applications this month and has started to recognise the patterns without needing any software at all. That’s harder to get around, and it’s something I went into in AI Humanizer Tools for Recruitment Outreach: What Gets Replies.

FAQ

Is there a 100% accurate AI detector?

Short answer: no. I’ve had Walter Writes score something I wrote myself at 40%, so I know from experience that false positives happen. All the main tools (Turnitin, GPTZero, Proofademic, Originality.ai) have this problem. They’re more reliable on raw, unedited model output than on anything that’s been worked over or edited. Use them as one signal alongside other things, not as a decision on their own.

What’s the best free AI detector for HR?

Depends what you need. Zero friction: aichecker.tech and aidetector.ac, both require no account. If you want more than a single number, Proofademic’s free trial gives you 1,000 words with sentence-level output, which is genuinely more useful than a percentage. GPTZero has a free tier too. None of the free versions are as thorough as the paid ones, but for occasional checks they’re fine.

Are plagiarism checkers 100% accurate?

No, and “plagiarism checker” covers a few different things depending on the tool. The ones people use most (Grammarly, Copyscape) are looking for text that matches something that already exists somewhere online. If a candidate writes a completely original AI-generated cover letter, those tools won’t flag it. AI detectors are working on something different: patterns in how the writing is structured. Neither approach is reliable enough to be the sole basis for rejecting someone.

What’s the best Turnitin AI detector alternative for recruitment teams with no budget?

Proofademic’s free trial covers 1,000 words with the full sentence-level breakdown, no card required. For a cover letter, that’s typically enough. For ongoing free use without limits, aichecker.tech and aidetector.ac are the most practical, though neither gives you the sentence-level detail. GPTZero’s free tier is limited on daily scan volume, so it works for occasional use but not regular screening.

There are tools that help. The catch is knowing which ones work for recruitment comms, not just for content teams or marketing.

The best AI writing tools for HR professionals prioritize tone control and humanization over output speed. Walter Writes leads this category because it rewrites AI-generated text at the structural level, not through word swaps, making it genuinely useful for the kind of writing recruiters do every day: job descriptions, rejection emails, candidate updates, and hiring manager communications where getting the tone wrong means complaints or drop-offs.

What recruitment writing requires

Most AI writing tools were built for marketing. Fast, punchy, persuasive. Not what I need.

Recruitment writing has to be clear without being cold. Professional without being robotic. A rejection email that sounds copy-pasted, even if it’s technically polite, gets screenshotted and posted somewhere. A job description full of buzzwords gets ignored by the people you’re trying to reach.

The tools I use regularly have to pass a few basic tests. Does the output sound like a person wrote it? Can I send it without doing another full rewrite? Would a candidate read this and feel respected rather than processed? Most AI tools fail at least one of those. A few don’t.

The best AI writing tools for HR professionals

1. Walter Writes - best AI writing tool

I recommend Walter Writes to anyone in recruitment or HR who uses AI for their day-to-day writing. Not because it has the most features. Because it solves the right problem.

When you paste a draft from ChatGPT or Gemini into Walter, it doesn’t swap synonyms or shuffle sentence order. It restructures cadence, breaks up predictable paragraph patterns, and removes the flat rhythm that makes AI writing obvious. For recruitment, that distinction matters a lot. A rejection email that reads like it was generated by a machine, even if the words are fine, still feels dismissive. The gap between “technically correct” and “human-sounding” is exactly where Walter operates.

Three rewrite strengths: Simple, Standard, or Enhanced. I run Standard on most things: rejection emails, job descriptions, candidate update messages. Enhanced when I’m starting from a particularly dense or jargon-heavy draft from a hiring manager who really went for it. The output consistently preserves meaning while improving tone, and ngl, that combination is harder to find than I expected. I wrote more about the full workflow for job posts specifically in this earlier piece if you want the detail.

The built-in AI detector is something I use more than I thought I would. After every rewrite, you get a score showing how the text is likely to register across major detection tools. That matters in recruitment more than people admit. If you want to make sure your content doesn’t read as obviously machine-generated before it goes out, having both functions in the same editor means no toggling between platforms. Most tools do one or the other. Walter does both. That alone makes it worth trying.

Which AI checker can also humanize text? Walter Writes. It comes up enough in HR communities that it’s worth stating directly.

Free 300-word trial, no credit card required. Starter is $8 a month on annual billing, 30,000 words per month, which is more than enough for most recruiters. There’s also a Teams plan at $99 a month for up to 10 members if you’re on a bigger team.

Try it at Walter Writes AI Humanizer

2. TextHumanizer.com

Quick, focused, almost no learning curve. That matters when you have a stack of rejection emails to send before end of day. Output quality is solid for shorter pieces: job posting tweaks, quick candidate updates, brief hiring manager notes. For longer or more nuanced communications, I’d want a manual pass before sending. For fast, low-stakes humanization, it earns its place.

3. HumaniseAI.ai

Built to remove AI patterns from text. For recruitment purposes that means softening anything that came out of a generation tool sounding stiff and impersonal, which is basically every first ChatGPT draft of a rejection email. The focus is tone transformation, not content generation, which is how most recruiters use AI writing tools anyway: to refine something that exists, not to write from scratch. Worth testing if your main need is making AI-generated drafts sound more human.

4. EssayHumanizer.ai

The name undersells it for professional use. EssayHumanizer.ai is particularly effective on text heavy with passive constructions and subordinate clauses, which is exactly what comes back when hiring managers write their own job descriptions and send them over expecting you to post them as-is. If your biggest challenge is transforming that kind of dense, formal draft into something a candidate can read, this one earns a place in the toolkit. Think of it as a corporate-to-plain-English converter. It doesn’t write for you. It just makes the text readable.

5. Jasper AI

Jasper has a solid template library and team collaboration features. Recruiters use it mostly for job descriptions and onboarding content where speed matters and emotional stakes are lower. The limitation: Jasper generates content rather than refining it. You’re starting with AI-written text that still needs a humanization pass before it goes to a candidate. Fine for volume content. For anything where tone matters, pair it with something that humanizes.

6. Copy.ai

Solid workflow features, easy to get started with. Useful for job postings and recruiting copy at volume. Like Jasper, it’s a generation tool, so output tends to be clean but still AI-patterned. Better suited for content pipelines than for individual candidate-facing messages where tone is the whole point.

7. Grammarly Business

Probably already in your company’s stack. Reliable for catching grammar and clarity issues. The tone detection feature can flag writing that reads too formal or too blunt, which is useful.

Just to be clear: Grammarly doesn’t humanize AI output. It edits what’s already there. If it’s your only layer on top of an AI draft, the output will still read as machine-generated. Use it at the end of the process, not the beginning.

Which AI writing tool is best for HR communications?

This comes up in recruitment communities fairly often. My answer depends on what someone is trying to do.

If you’re refining AI-generated drafts for anything candidate-facing, you need structural rewriting, not surface-level cleanup. That’s where Walter Writes separates from everything else I’ve tried. Most basic humanizers handle synonym replacement and basic phrasing. Walter works at the level of cadence, rhythm, and meaning preservation, which matters when the message is a rejection, a timeline update, or a job description that needs to reach the right people. I ran a comparison test on outreach messages a while back and wrote up the results here. The reply rate difference was enough to make it my default.

If you only need grammar help, Grammarly is enough. If you need AI-generated text to sound like it was written by someone who respects the reader, you need something that works deeper than that.

How to humanize AI content without losing the message

Not about making AI writing undetectable for its own sake. It’s to get to something sendable without spending an hour on every draft.

My process is straightforward. Start with the key info I need to communicate. Use AI to get a rough draft. Run it through a humanization tool. Then do one final pass focused on tone, asking myself: would I be comfortable if a candidate screenshotted this and posted it somewhere?

That loop gets me to a usable draft in under twenty minutes for most routine communications. For anything where wording really matters, like a candidate who’s been waiting weeks and is probably frustrated, I take a few extra minutes to read it aloud. If it sounds stiff, it reads stiff. I covered more of this workflow in my Walter Writes review if it’s useful.

Frequently asked questions

Which humanize AI is best for HR writing?

I get this question fairly often, and my answer hasn’t changed: Walter Writes. It works at the structural level rather than just swapping words around, which for recruitment writing is the difference between a message that feels human and one that feels like a mail merge. The built-in AI detector is useful too, especially if you want to check how the output reads before it goes out.

What is the best AI tool to humanize text for professional communications?

Depends what you’re trying to do. For general business writing, several tools are fine. For recruitment, where “professional” also has to mean “won’t make a candidate feel like a number,” Walter Writes is where I land. The adjustable rewrite strength matters here. Standard handles most day-to-day writing. Enhanced handles denser or more formal source material.

Can an AI checker also humanize text?

Yes, and Walter Writes is the clearest example I’ve found. After every rewrite, it scores the text against major detection tools including GPTZero and Turnitin. For anyone producing recruitment content or HR documentation, having both functions in one editor is genuinely useful. No switching between tabs.

How do I humanize AI content for free?

Walter Writes has a free 300-word trial with no credit card or login required. For shorter communications like a rejection email or a quick candidate update, that’s enough to test the output quality. Paid plans start at $8 a month on annual billing.

Recruitment writing is one of those areas where tone mistakes have visible consequences. A rejection that sounds robotic gets shared. A job description that reads like a buzzword generator gets ignored. A candidate update that feels like a form letter makes people drop out.

The tools above reduce that risk. Walter Writes is where I’d start if you’re using AI to draft anything that goes to candidates or hiring managers. The difference between raw AI output and a properly humanized draft is the difference between something that processes people and something that respects them.

No one flagged it. No one complained. But I knew. The phrasing was too smooth. The transitions too tidy. It read like a press release, not like something a person had written to another person.

These days I write almost everything with ChatGPT as a first draft and an AI humanizer to clean it up afterwards. Not because I’m lazy, but because I’ve got about 30 open roles across four departments, I’m the only person handling candidate comms, and the hiring managers take three business days to return feedback when it’s good. I don’t have time to write every rejection email from scratch. No one does.

The best free AI humanizer I’ve found for this kind of work is Walter Writes. I’ve tried others. This is why I stuck with this one.

Walter Writes is the best free AI humanizer for professional writing tasks like job descriptions, outreach emails, and candidate communications. It rewrites AI-generated text at the structural level rather than just swapping synonyms, preserving the original meaning while varying sentence rhythm and length. The free tier gives you 300 words per use with no login or credit card required, which covers most short-form recruitment comms without needing a paid plan.

Why Recruitment Writing Gets Flagged as AI

Job descriptions and candidate emails are two of the most generic document types in professional life. They follow the same structure every time. “We’re looking for a motivated and enthusiastic professional to join our growing team...” followed by a bullet list of requirements copy-pasted from a similar role on LinkedIn, finished with something about competitive salary and excellent benefits.

When you ask ChatGPT to write one of these, it defaults to exactly that pattern. Because that pattern is everywhere in its training data.

AI detectors, which are increasingly common in ATS (Applicant Tracking System) platforms and which plenty of candidates run their correspondence through, pick up on these patterns quickly. High perplexity scores, low burstiness, sentences that are all roughly the same length with the same confident cadence. It doesn’t matter that the content is accurate. It reads like it was generated, because it was.

The same problem applies to outreach messages. Candidates can tell when they’re receiving a templated, AI-written InMail. The phrasing is slightly off. Too polished. Too even. Too similar to the last 15 messages they already ignored.

This isn’t just an aesthetic problem. A job post that reads as AI-generated gets lower engagement. Candidates scrolling LinkedIn move past it. The ones who do apply sometimes assume the whole company communicates this way and question the culture. For junior recruiters like me who don’t control the brand or the salary budget, candidate experience in the communication itself is one of the few things we own.

What I Need From a Free AI Humanizer

Before I landed on Walter Writes, I tried a few others. Quillbot has a humanizer but it felt like it was mostly paraphrasing. The output was still too tidy, too smooth. One or two other tools either broke the grammar, changed the meaning in ways I then had to fix, or were only free for about 20 words before pushing me to a subscription.

What I needed was something specific:

• Works on short to medium text: a paragraph, an email, a job description section

• Doesn’t introduce mistakes I have to go back and clean up

• Keeps the meaning intact, especially for job descriptions where the requirements can’t change

• Has a free tier that’s genuinely usable, not a teaser

• Produces something that sounds like a person wrote it

That last one is harder to hit than it sounds. A lot of humanizer tools just shuffle the words around. The result is different text, but it still has the same flat rhythm, the same even confidence. What you need is variation. Sentences of different lengths. Paragraphs that don’t all follow the same structure. That kind of texture is what AI detectors are measuring, and it’s what most cheap tools miss.

Why Walter Writes Works for Recruitment Copy

The thing that stood out about Walter Writes Humanizer compared to what I’d tried before is the rewrite strength. You can choose between Simple, Standard, or Enhanced depending on how much restructuring the text needs.

For outreach messages I usually use Standard. The text is already fairly close to what I want, I just need the robotic edge taken off. For longer job descriptions that were written entirely by ChatGPT and still sound like it, Enhanced does a more thorough job of breaking up the structure.

The meaning preservation is the part that matters most for recruitment work. A job post has very specific information in it. Salary range, seniority level, reporting line, location, required experience. If a humanizer scrambles those details, I’ve got a real problem. Walter Writes adjusts the tone and rhythm, not the facts.

There’s also a built-in AI detector inside the same editor. After every humanization, I can see the AI likelihood score without copying the text into a separate tool. Small thing, but it saves time when you’re doing this 10 or 15 times a day.

The free tier is 300 words per use, no login, no credit card. That covers most short-form communications without needing a subscription. Rejection emails are usually 100 to 150 words. Candidate update messages, similar. If I’m working on a longer job description, I split it into sections and run each one separately.

My Actual Workflow for Writing Job Posts

Here’s how a typical job description ends up on our jobs page or LinkedIn.

I start with a brief from the hiring manager. This is usually a half-formed list of requirements and a paragraph about the team that doesn’t quite make sense. I feed that to ChatGPT with a prompt asking for a structured job description in the format we use.

ChatGPT gives me a solid draft in about 30 seconds. It’s accurate and well-organised, but it’s also very obviously AI-written. Every sentence is confident and smooth in a way no actual job post written by a real recruiter is. Real job posts have inconsistencies and slightly off phrasing. This one is too clean.

I paste it into Walter Writes and run it on Enhanced. The output restructures the sentence rhythm, breaks up some of the longer uniform paragraphs, and introduces enough variation that the whole thing reads differently. I check the AI score after and it’s usually below 20%.

Then I give it a final read and fix anything that doesn’t sound right or doesn’t match how we’d describe the role. This whole process takes about 10 minutes. Compared to 45 minutes if I’m writing from scratch, or the 20 minutes of manual editing I used to do trying to de-AI ChatGPT output by hand, it’s a meaningful difference.

For outreach messages and rejections the same process applies, just shorter. Five minutes per piece once you’re used to it.

The Free vs Paid Question

The 300-word free limit covers most short-form communications. If you’re a recruiter writing a handful of emails a day and cleaning up the odd job description, you’ll probably get a decent amount done on the free tier before hitting any wall.

For higher volume work, the Starter plan is $96 per year, which is $8 a month. That comes with 30,000 words per month. For most individual recruiters handling a reasonable number of open roles, that’s more than enough. There are higher tiers if you’re doing more.

I haven’t upgraded yet, but the path to the value making sense is pretty clear if my workload increases. Right now the free tier does what I need. It’s the most genuinely usable free AI humanizer I’ve come across for professional writing.

A Quick Check Before I Send

I’ve started running anything that goes to candidates through a quick AI detection check before it leaves my outbox. Not because I think anyone’s going to penalise me for using AI-assisted writing. I just want to know it reads as human before it reaches someone who’s hoping to hear from a real person.

For that final check I use Proofademic’s free detector. It gives a sentence-level breakdown showing which parts are flagging as AI-generated and which aren’t. That’s useful when I want to see exactly what’s still sounding off rather than just a single overall score. Knowing it’s sentences three and seven, rather than just “partially AI,” gives me something to fix.

Between the humanizer and that final check, I’ve basically stopped worrying about whether my candidate comms sound like a robot wrote them.

Frequently Asked Questions

What is the best AI humanizer for free? Walter Writes is the best free AI humanizer I’ve used for professional writing. The free tier gives 300 words per use with no login or credit card required, and the rewriting is structural rather than just synonym-swapping, which makes the output sound more naturally human.

Does a free AI humanizer work well enough for job descriptions? For most job description sections and short candidate emails, yes. The 300-word free tier on Walter Writes covers the kind of short-form comms that make up most of a recruiter’s day. For longer documents you can split text into sections and run each separately, or consider a low-cost paid plan.

Can AI detectors tell the difference after humanizing? Depends on the tool. Basic humanizers that only paraphrase often still produce detectable patterns. Walter Writes rewrites at the structural level, varying sentence length and rhythm, which tends to score lower on AI detectors. I usually check after with the built-in detector and aim for below 20% AI likelihood before sending.

Should recruiters be worried about using AI humanizers? Most companies haven’t set explicit policies on this yet. The practical concern isn’t legal, it’s relational. Candidate experience suffers when communications feel automated. The point of humanizing isn’t to hide AI use, it’s to make sure the output sounds like a person is behind it, which is what candidates and hiring managers reasonably expect.

The writing was clean. Properly structured. Each opening paragraph hit the same beats: excited about the company, relevant background, ready to contribute. Every sentence was well-formed and nothing felt hasty or unrefined.

At first, I couldn’t figure out what I was seeing. Then it clicked. It was the exact feeling I get reviewing my own first drafts from ChatGPT before I’ve edited them. That slightly too-smooth, slightly too-even quality that reads professionally but doesn’t quite sound like a person wrote it.

I use AI tools myself. I’m not judging candidates for doing the same. But the sheer number of applications that read this way has changed how I screen, and from what I can tell from conversations with colleagues, I’m not the only one noticing.

So the question a lot of candidates are asking is fair: does Turnitin detect AI in cover letters? Turnitin has AI detection built in, but most employers aren’t running cover letters through it. What’s happening is more human than that. Recruiters who read hundreds of applications have started developing pattern recognition for AI-generated writing, and some organisations are using standalone tools like GPTZero or Originality.ai to flag it. The risk for most candidates isn’t a system score. It’s a person who reads enough to notice.

What Turnitin is and how it’s used in hiring

Turnitin was built for universities. Its primary purpose is plagiarism detection in student submissions, and in recent years it added an AI writing detection feature that analyses text for signals associated with AI-generated content.

In academic contexts, a Turnitin AI score can affect grades and trigger formal misconduct reviews. That’s the environment it was designed for.

Most recruitment teams don’t use Turnitin. It’s an academic licensing product and there’s no standard HR workflow that routes cover letters through it. For entry-level roles, admin positions, and most general hiring, it’s very unlikely to come up.

Some sectors are starting to bring AI detection into their process, though. Graduate schemes, law firms, financial services recruiters, and communications-heavy roles are more likely to run some form of screening. Not always Turnitin specifically, but tools that do the same job. GPTZero and Originality.ai come up regularly when colleagues talk about this.

For the vast majority of applications, the real screening is happening in a person’s head.

What recruiters are noticing

I can only speak from what I’ve seen, but I think it lines up with what a lot of people in recruitment are experiencing right now.

AI-generated cover letters have a specific texture. The structure is almost always the same: an enthusiastic opening about the company, a middle section walking through relevant experience with clear transitions, and a closing paragraph about the value the candidate would bring. Grammatically correct throughout. Nothing out of place.

What’s missing is the messiness that comes with a real person writing. Uneven sentence lengths. A slightly unexpected word choice. A moment where someone just said what they meant rather than what sounded impressive. Even the best cover letters from real people have something slightly imperfect about them, and that imperfection is reassuring.

This connects to something called perplexity in AI detection terms. Perplexity measures how predictable text is at the word-by-word level. AI models choose words based on statistical probability, which makes the writing smooth and coherent but also flat and uniform. Human writing, even polished human writing, naturally introduces more variation.

Recruiters screening high volumes haven’t read a paper on perplexity. But they’re responding to the same thing the detectors are measuring. It just registers as “something feels off about this” rather than a percentage score.

How AI detectors analyse text

For anyone who wants to understand what they’re up against, here’s a basic picture of how AI detection tools work.

Most tools, including Turnitin’s AI detector, GPTZero, and Originality.ai, are looking at two main signals. The first is perplexity: how predictable the word choices are, sentence by sentence. The second is burstiness: how much the sentence length and complexity varies throughout the text. Human writing tends to be high on both. AI output tends to be low on both.

These tools aren’t infallible. False positives exist, meaning genuinely human-written content can get flagged, particularly formal academic writing or content that follows structured conventions. The technology is improving but it’s not a definitive verdict.

What they’re good at is flagging unmodified AI output. If someone generated a cover letter with ChatGPT and submitted it without significant changes, those statistical patterns are usually detectable. The more raw the output, the clearer the signal.

Can an AI humanizer bypass Turnitin?

This is what most people are really asking. And the answer depends on what the humanizer does.

Surface-level paraphrasing, which is what basic tools do, usually doesn’t help much. If you’re just swapping synonyms or slightly rearranging sentences, the underlying statistical fingerprint is still there. Turnitin and similar tools are looking at patterns across the whole text, not individual word choices.

What does help is structural rewriting. That means changing how ideas are expressed at the sentence level, varying rhythm, mixing short and long sentences, rephrasing rather than paraphrasing, and generally disrupting the uniform cadence that AI naturally produces.

I started using Walter Writes AI Humanizer for my own work output, mostly for rewriting job descriptions and candidate emails. The difference from basic paraphrasing tools is noticeable. It restructures sentence patterns rather than just swapping words, and it has adjustable rewrite levels depending on how much transformation the text needs.

There’s a built-in AI detector included, so you can check your AI-likelihood score right after rewriting without switching between separate tools. Tested results on Turnitin put it at 100% Human after humanization. GPTZero and Originality.ai scores come out similarly. That lines up with what I see in practice: text that’s been properly restructured reads differently and scores differently.

Does Turnitin detect Walter Writes AI output?

A few people have asked this directly, so it’s worth being clear.

Turnitin’s detector is looking for statistical patterns in the text. If that text has been processed through a humanizer that does structural-level rewriting rather than just synonym substitution, those patterns are disrupted. Walter Writes is specifically designed for this: it removes the fingerprints that detectors are trained to find, not by disguising them, but by rewriting in a way that changes the underlying structure.

The before/after scores on the Walter Writes site show Turnitin going from 95% AI to 100% Human. Individual results may vary depending on the original text and rewrite settings. But the core mechanism works because it’s addressing the right problem.

What candidates are getting wrong

The mistake isn’t using AI. It’s using AI and submitting the first draft.

Unedited ChatGPT output has tells beyond the statistical ones. The enthusiasm sounds exactly like every other applicant’s enthusiasm. The experience section is generic and could apply to any employer. There’s nothing specific in it: no real reason for wanting this company, no concrete example that only this person could give, no observation that shows they’ve been paying attention.

When a cover letter has that kind of specificity, it reads differently. Even if the structure came from AI, the details are yours. That combination of efficiency and personal content is what works.

Running the output through a humanizer like Walter Writes and then editing in specifics is a sensible workflow. The humanizer handles the AI detection risk. The editing makes it relevant.

What I’m looking for when I read a cover letter isn’t proof that someone struggled to write it unaided. I’m looking for evidence that a real person thought about this application. Those two things aren’t the same.

Frequently asked questions

Does Turnitin detect AI in cover letters?

Turnitin has AI detection capability, but it’s an academic licensing tool and most employers don’t use it to screen cover letters. Standalone detectors like GPTZero and Originality.ai are more common in hiring contexts. For most roles, the main risk is a human reviewer noticing AI patterns rather than a formal detection score.

Is humanized AI text detectable by Turnitin?

Text that’s been surface-level paraphrased often still gets flagged because the underlying statistical patterns, specifically perplexity and burstiness scores, remain similar to the original. Text that’s been structurally rewritten at the sentence level is much harder to detect because the signals those tools measure are disrupted.

Can a recruiter tell if AI wrote a cover letter?

Not with certainty, but recruiters who read a high volume of applications develop a feel for AI-generated writing. The uniformity of structure, the absence of awkward phrasing, and the generic enthusiasm are patterns that stand out once you’ve read enough of them. It doesn’t lead to automatic rejection in most places, but it does make the application less memorable.

Can AI humanizer output be detected?

It depends on how the humanizer works. Basic paraphrasers don’t do enough because the statistical fingerprint is still present. Humanizers that do structural-level rewriting, adjusting sentence rhythm, cadence, and pattern, are much harder to detect. That’s the distinction between tools that swap words and tools that do deeper structural work.

What’s the best approach for AI-assisted cover letters?

Use AI to draft, run it through a humanizer, then edit in specific details that only you could provide: the genuine reason you want this role, a real example from your background, something that shows you’ve paid attention to this particular company. That combination handles both the detection risk and the “sounds like everyone else” problem.

That’s normal. Average reply rates for recruiter cold messages on LinkedIn sit somewhere between 10-15%, and that’s for people doing it well. I wasn’t doing it well.

My problem wasn’t that I was messaging the wrong people. It was that my messages sounded exactly like every other message candidates were already getting. Professional. Polished. Forgettable.

I’ve been in entry-level recruitment in the UK for a few years now and the biggest part of the job is communicating with candidates. Getting replies is basically the whole game. If nobody responds, the pipeline dries up, the hiring manager gets frustrated, and your week goes sideways. I’ve spent more time than I’d like to admit figuring out what actually works.

Writing recruiter outreach messages that get replies comes down to three things: the opening hook, how relevant the message feels to that specific person, and a tone that sounds like you took 30 seconds to actually think about them. Most templates fail because they feel like templates. Short, specific, and personal beats long and thorough every time.

Why most outreach messages go unanswered

Most candidates, particularly passive ones, get a lot of recruiter messages. They’ve seen the “exciting opportunity” opener. They’ve seen “I came across your profile and was really impressed.” They’ve seen messages that are basically copy-pasted job descriptions with their name dropped in at the top.

When a message looks like a template, the brain processes it as noise. It doesn’t matter if the opportunity is genuinely right for them. First impressions in an inbox happen in under three seconds, and most recruiter messages don’t survive that initial scan.

The other issue is length. There’s this instinct in recruitment to be thorough, to explain the role, cover the benefits, mention the company. But a cold outreach message isn’t the place for that. Your only job is to get a response. The details come later.

A former manager once told me that a long message signals low confidence in your pitch. I didn’t fully agree at the time, but now I think she was right. If your opening message needs three paragraphs to be convincing, it probably won’t be.

What makes someone actually reply

I’ve tested a lot of approaches. The ones that actually generate responses share a few things.

First, they reference something specific. Not fake-specific like “your background in X is impressive” where X is just their job title. Actual specific: a recent career move, a niche skill they’ve listed that’s unusual, or something that suggests they might be at the right point in their career to consider a move.

Second, they’re short. My highest-response messages are usually under 80 words. Sometimes under 60. When I trimmed a message from 200 words down to 65, the reply rate went up noticeably. That surprised me at first, but it makes sense. Short messages feel considered. Long ones feel automated.

Third, they ask for exactly one thing. Not “would you be open to a chat, and if so what times work, and could you also have a look at this JD.” One thing. Usually something low-commitment like “would it be worth a quick 10-minute call?” or “happy to send more details if that sounds interesting.”

That last framing is something I use a lot. It puts the decision entirely in their hands, which tends to reduce the friction of replying.

Templates I actually use

Here are a few formats that have worked for me. Don’t copy these exactly. Use them as a starting point and adjust them for whoever you’re contacting and what you actually know about them.

For passive candidates with a relevant but niche skill:

Hi [Name], spotted your background in [specific skill]. We’re currently building out that side of the team and this role isn’t advertised yet. No pressure at all, but happy to share details if it sounds worth a look.

For someone who recently changed roles:

Hi [Name], saw you recently moved into [new role]. Timing might be off, but we have something in [adjacent area] that might be interesting further down the line. Would it be okay to stay in touch?

For a referred candidate:

Hi [Name], [mutual contact] thought it might be worth us speaking. We’re hiring for [role] and it seemed relevant given your background. Let me know if you’d want to hear more, no rush but happy to send over details.

For someone with strong tenure in a related area:

Hi [Name], noticed you’ve been focused on [area] for a few years now. We’re looking for someone with exactly that kind of experience for a role we’re filling at the moment. Would a quick call make sense?

The through-line in all of these is the same: short, direct, one ask, no pressure framing. The “no rush but” and “happy to tweak” language is genuine. It’s not filler, it signals this isn’t a mass send.

The AI problem I didn’t expect

I started using ChatGPT to help draft outreach messages about a year ago. It saved time at first. I’d describe the role, describe the candidate’s profile, and ask for a draft.

The drafts were fine. Grammatically correct, professionally worded. But they had this quality to them, a slightly too-clean, slightly too-formal thing, that made them feel corporate in a way that wasn’t landing. Phrases like “I wanted to reach out to discuss a potentially compelling opportunity” would come back and I’d spend as much time editing those out as I’d have spent writing from scratch.

The issue is that candidates can feel the difference between a message a human wrote and one that came out of a machine. Hard to explain exactly, but the pattern of language is just slightly off. Uniform sentence length. Certain transition phrases. A kind of over-politeness that doesn’t sound like a real person.

I tried a few tools that claimed to make AI text sound more natural. Most just shuffled words around or swapped synonyms. That didn’t fix the underlying pattern. The message still read as AI-generated to anyone paying attention.

To directly answer the question of whether an AI humanizer can be detected: yes, some of them can, especially the ones that only do surface-level edits. What actually works is restructuring the sentence rhythm and varying the phrasing patterns, not just swapping words.

What actually fixed the tone problem

The tool I use now is Walter Writes. I draft the initial message in ChatGPT, or sometimes write a rough version myself, then run it through Walter Writes to get the sentence structure and rhythm closer to how a person actually writes.

The difference isn’t dramatic on every message, but it’s consistent. Walter Writes restructures phrasing at a sentence level rather than just swapping words. The result reads like something I’d actually type, not like something pulled from a corporate LinkedIn outreach template.

The reply rate difference is noticeable. I haven’t tracked it precisely, but messages that go through a humanizing pass before I send them feel less like form letters. They read like I wrote them, which is sort of the point.

It also takes some of the anxiety out of sending. I already second-guess the tone of every message I write. Having something that makes the output feel more natural before it goes out helps a lot.

For anyone working on an entry-level budget, the free tier on Walter Writes covers up to 300 words per session. That’s more than enough for a standard cold message. So if you’re asking whether there’s a free AI humanizer that actually works for recruiter messages, the answer is yes, and that’s where I’d start before committing to anything paid.

Questions I get asked most about recruiter outreach

Does using AI to draft outreach messages count as being dishonest?

I’ve thought about this more than I’d like to admit. The message still represents your actual role, your actual company, and your actual intent to speak to that person. Using AI to help draft and then editing it is no different from using a template or asking a colleague to review your wording before you send. The core communication is still yours.

Can candidates tell when AI wrote a recruiter message?

Sometimes, yes. Especially if the message is too clean, uses certain giveaway phrases, or has that flat sentence rhythm AI tends to produce. This is exactly why the humanizing step matters. A message that reads like a real person wrote it is much less likely to be mentally filed under “mass outreach.”

How do you humanize AI text properly for professional messages?

Don’t just paraphrase or swap words. Use a tool that rewrites sentence structure and varies the rhythm. Then read it back out loud. If you wouldn’t actually say that sentence, change it. The best humanized output is the kind you can read back and genuinely believe you wrote yourself.

What if I write a good opener but then have to actually speak to them?

This is the part that matters most. The outreach message is just the door opener. Once someone replies, the conversation is yours to carry. If you referenced something specific in their profile, you need to actually know that detail when you speak to them. Don’t fake specificity you can’t back up.

What’s the best way to use AI when sending multiple outreach messages at once?

Batch drafting works well. I’ll sometimes draft five or six versions in one ChatGPT session using slightly different prompts for different candidate profiles, then humanize each one individually. It speeds things up without making the messages feel identical.

What I’d change if I were starting over

The biggest shift in my approach isn’t the templates themselves. It’s how much I think about tone before any message goes out.

Candidates are people getting a lot of noise in their inboxes. Anything that makes your message feel like it was written for them, by a person, works in your favour. That means keeping it short, keeping it specific, and making sure the language doesn’t sound like it was produced in bulk.

If you’ve been struggling with reply rates, start with length. Cut your messages in half and see what happens. Then look at personalisation. Then look at tone.

Let me know what’s been working for you, or what type of outreach situation you’d like me to cover next. Happy to share more.

I’d written it, read it back, hated it. Closed the tab. Came back later, hated it again. Eventually just sent it at five past five because I needed to leave.

If you work in recruitment, especially at the junior end where it’s basically you, an ATS you’re still figuring out, and a hiring manager who replies to emails twice a week if you’re lucky, you probably know that particular feeling. Sending something you’re not happy with because the alternative is staying late to agonise over twelve words.

I’ve been in this role just over a year now and the rejection email problem has genuinely taken up more of my mental energy than probably anything else. It sounds ridiculous. But getting the tone right on a rejection is harder than it looks, and getting it wrong has actual consequences.

Why AI rejection emails are so obviously AI

So, about eight months ago I started using ChatGPT to draft these. I thought it would save me the agonising. Type in the context, get back a usable paragraph, send. Done.

First attempt: “We regret to inform you that after careful consideration of your application, we have determined that your profile does not align with the current requirements of this role.”

Nobody says that. I kept staring at it thinking, has anyone in the history of employment ever actually said “your profile does not align with the current requirements” to a real human person? It reads like a rejection generated by an HR compliance bot in 2009.

The problem with raw AI output for this kind of thing isn’t factual accuracy. It gets the facts right. It’s that the writing has no texture. It’s uniformly formal in a way that feels completely inhuman. Every sentence the same weight, the same register. No imperfections. No warmth. Candidates can tell, they’ve seen enough of these to recognise when something came from a template, and when they can tell, the rejection stops being just a no and starts feeling like evidence that the company didn’t bother.

That’s a reputational issue for the whole employer brand, but in practical terms it also just makes me feel bad every time I send one. I know what it’s like to get those emails from the other side.

So I started actually digging into how to humanize AI text in a way that would work for professional writing specifically. Not just grammar corrections. Something that would change the actual feel of the output.

What I tried first (and why it didn’t quite work)

I gave Grammarly a proper go for a few weeks. It’s a solid tool for what it does, but it’s really a grammar and style checker, not a tone converter. It’ll tell you if you’ve used passive voice too much or if your sentences are too long. It won’t tell you that your rejection email sounds like it came from a legal team. Those are different problems.

QuillBot I tried for maybe a week. The paraphrasing is good. But paraphrasing and humanizing aren’t really the same thing. Paraphrasing changes the words while keeping the structure and register. What I needed was something that would change the feel of the writing: the rhythm of it, how it lands when you read it. That’s a different level of intervention.

ngl I also tried just manually rewriting everything for a bit, which was just me spending twenty minutes on emails that should take five. Reading things back out loud, tweaking, still not being happy, sending them anyway. Not a real solution.

How I actually do this now

This is going to sound obvious but it took me a while to land on it.

I use ChatGPT for the content skeleton only. Role title, the actual decision, what the next step is. Nothing else. No filler phrases, no softening language. I strip all of that out because I’m going to be adding something better back in. The AI draft is raw material, not something I’d send.

Then it goes through Walter Writes. What the humanisation layer does is change the rhythm and register of the text rather than just substituting words. Contractions come in. The overly even sentence length breaks up. The policy-document formality drops off. It’s not making every decision for me, but it’s handling the baseline in a way that would otherwise eat twenty minutes of my day.

What I’m left doing is the judgment calls that no tool can make: what does this specific email actually need? A rejection after a quick first-call screen is a different email to one after someone’s done two stages and submitted a task. The investment is different. The emotional weight is different. Having the language baseline handled means I’m making one or two of those judgment calls instead of thirty.

I also specifically look for what I’d call false warmth. Some AI humanizers overcorrect and start adding phrases like “Your passion genuinely shone through” or “We were really impressed by your background.” That’s actually worse than robotic because candidates can read performative sympathy. I cut anything like that before it goes out.

The emails I send are almost always three short paragraphs now. Paragraph one: the decision, stated directly. Paragraph two: something genuine and specific about what the candidate did or how far they got. Paragraph three: next steps, or an invitation to apply again only if that’s true. Clean, predictable structure. Candidate finds the answer without having to read through layers of softening first.

The actual principles that make a difference

Took me a while to work these out. HR resources are not short on interview guidance. The post-interview communication is another matter.

The biggest one is specificity. “After reviewing a number of strong candidates” is meaningless and candidates know it. “We had a competitive shortlist and this one came down to a specific technical requirement” at least gives the rejection some actual shape. It’s not oversharing internal process. It’s just not hiding behind language so generic it could apply to literally anyone.

Say sorry once. One sentence. More than one and it reads as either anxious or performative. Acknowledge it, move past it.

Match the email to the investment the candidate made. Someone who applied through a job board and had zero contact with anyone doesn’t need three paragraphs. Someone who came in for two interviews and did a four-hour task deserves more than a form letter. This should be obvious but when you’re sending thirty in a day it requires conscious effort.

And please, stop writing “we’ll keep your details on file” if you’re not going to look at that file again. Candidates remember. It’s one of those small broken promises that adds up to the general impression that recruitment departments can’t be trusted.

The humanise AI text step matters for all of this because AI drafts tend to pile on politeness and professionalism until the email says nothing at all. The humanising step is what strips it back to something actually usable. Without it, the principles above get buried.

Tbh, the bar I’m actually aiming for is that the email should read like someone spent five minutes on it and cared about those five minutes. Not thirty seconds, approved, next. At volume, that genuinely takes tool support.

Being honest about scale

This whole thing works better in theory than in practice at scale, and I want to be straight about that.

When I’m sending thirty rejections in a day, I’m still working from a consistent base structure. The format doesn’t change. What changes is the middle paragraph: what’s specific to this person, this role, this stage, the one honest thing I can say about why we’re not progressing. Everything else stays consistent out of necessity.

The question of how to humanize AI text at scale is really about deciding where to put the variable effort. Mine goes into the specifics. Walter Writes handles the baseline conversion from stiff ChatGPT draft to something that sounds like a person typed it. I add the specific detail on top.

Between those two things, what I’m sending is better than what I was producing when I was writing everything from scratch while also dealing with interview scheduling, candidate questions, and the perpetual task of chasing hiring managers who’ve apparently forgotten what email is. Junior recruitment does not come with long uninterrupted blocks of writing time.

If you’re in a similar situation and you’ve been searching for an AI humanizer for professional writing, the thing to look for is texture change rather than vocabulary change. Tools built for paraphrasing produce writing that reads differently on the surface and identically in terms of feel. The mechanical quality is still there, just dressed differently. What you need is something that genuinely changes the rhythm.

I can’t claim this has fixed everything. But I get fewer one-word replies. More responses where someone thanks me and says they’ll keep an eye on future roles. Those weren’t happening much before.

Maybe that’s not how anyone else measures it. It’s how I’m measuring it. Let me know if it’s useful or if you’ve got a completely different approach. Genuinely curious what other people at this level are actually using.

About a year ago I started leaning on ChatGPT for the drafts. It’s fine. Gets me somewhere close to what I need and saves a chunk of time. The problem is the output always reads a bit off. Too formal. Slightly stiff. Like it was written by someone who’s studied a lot of HR documents but hasn’t actually sent a rejection email to a real person before. I had a candidate reply once with “thanks for the automated message” and I’d genuinely spent ten minutes on that email. That one stuck with me a bit.

So I started looking at AI humanization tools. Tested a few over the past few months. Walter Writes is the one I’m still paying for, which I think says something.

Here’s my honest take on whether it’s worth it.

Walter Writes is worth paying for if you’re producing a lot of AI-generated text that needs to stop sounding like AI. For recruiters specifically, it comes down to how much time you’re losing cleaning up raw output. If you’re sending 20+ messages a day that started as ChatGPT drafts, the gap between “slightly off” and “reads like a person wrote it” matters more than you might think. I consistently get usable text back in under 30 seconds. That’s not nothing.

What I actually use it for

The use case for a humanization tool in recruitment isn’t obvious on the surface. I’m not submitting anything to Turnitin.

But the problem is real regardless. Raw AI output has a quality to it that people notice even if they can’t name it. There’s this corporate flatness, phrases that nobody actually uses out loud. “We appreciate your interest in this opportunity” being the classic example. Nobody on our team says that. Ever.

Rejection emails are where I use it most. Writing 15 of them that each feel like an actual person wrote them is tedious in a way that gets old quickly. ChatGPT gives me a template to start from. Walter strips out the corporate layer and gives me something that reads like I was actually thinking about the candidate when I typed it. It’s a small thing but candidates notice.

JD rewrites are the other big one. Hiring managers send me rough drafts full of buzzwords and things like “must be comfortable working in a fast-paced environment” (which doesn’t tell anyone anything). I clean those up in ChatGPT first, then run the output through Walter to make it sound like a real team describing what the job actually is. I’ve noticed candidates ask better questions in interviews since I started doing this. That matters.

Follow-ups to candidates who’ve gone quiet are the third thing. These are hard to get right on your own. Too pushy and you damage the relationship. Too passive and nothing happens. Having a middle-draft to work from speeds that up.

Does Walter Writes actually work?

I was skeptical before I tried it, tbh.

Short answer: yes. It does work. The caveat is that “works” needs defining. If you mean “does it remove AI tells from text so it reads naturally,” then yes, it does. The output genuinely reads differently after processing. Sentence lengths vary. The stiff phrasing breaks up. It’s not just swapping words around, the structure of how ideas get expressed changes.

That said, it’s not perfect every time. About 80% of what comes back is ready to send or close to it. The other 20% needs another couple of minutes. But that’s still better than rewriting a rejection email from scratch, which I was doing regularly before.

I’ve run ChatGPT output through it, Claude output, older GPT drafts. Handles them all consistently.

The built-in AI detector is something I use constantly now and didn’t expect to. After every rewrite, I can check the AI likelihood score without leaving the editor. With the free tools I tried before, like Humanize.ai, I had to copy the output into a separate detector. Having both in one place saves time in a way that adds up over a week.

Walter Writes pricing: worth it on a junior salary?

I need to be straight here because I’m not someone with a work expenses card.

I’m entry-level. Subscriptions are a real conversation I have with myself. Before I paid for anything I went through the free tier to understand what I was actually getting. Walter Writes has a 300-word free trial, no credit card needed. That’s enough to test it properly on something real, which I did.

The Starter plan works out to $8 a month. 30,000 words per month. For context, that’s far more than I’ll ever use for recruitment comms. A rejection email is 60-80 words. A JD rewrite might be 300. If I’m processing 30 to 40 pieces a week I’m nowhere near that ceiling.

The question I kept coming back to was: what’s my time worth on a Monday morning? If I’m saving 30-40 minutes a week, that’s an easy $8. It’s a more honest way to think about it than comparing it to a Spotify subscription.

I used Undetectable.ai before this. Output was inconsistent and it wasn’t meaningfully cheaper. Walter Writes has more control over tone, which for recruitment specifically makes a real difference.

Can Walter Writes be detected?

This comes up a lot online. Worth addressing.

For my use case, the concern is different from an academic one. Nobody’s running my rejection emails through GPTZero. What I care about is whether the text reads naturally enough that candidates don’t clock it as automated. That’s the actual problem.

But on the detection question properly: Walter Writes Humanizer has three rewrite strength levels. Simple, Standard, Enhanced. Enhanced does heavier structural work and scores lower on AI detection tools. For the short-form messages I write in recruitment, Simple or Standard is usually enough.

The built-in detector gives a score after each rewrite. From what I’ve seen across my own outputs, Enhanced brings AI likelihood scores down a lot. The published data on the Walter Writes site shows before-and-after scores against GPTZero, Turnitin, Originality, and Copyleaks. My experience holds up with those results for the kind of text I’m running through it.

More importantly in practice: text that reads naturally doesn’t set off the instinct in people that something’s off. That’s the thing I’m actually solving for.

Is Walter Writes legit? My honest review after three months

I’m not the target user. The tool is aimed at students and content writers, not a junior recruiter in Stockport trying to get through their email queue before lunch.

But it fits what I need well enough that I’ve kept paying for it. The Chrome extension is the feature I use most daily. I can run a rejection email through the humanizer from inside Gmail without switching tabs. That kind of workflow efficiency is worth more than it sounds.

Things I actually like: the tone selector. You can pick casual, brand-safe, academic, or journalistic. I use brand-safe for most recruitment content and it consistently hits the right register. The speed is accurate too, under 30 seconds every time I’ve used it. And having the humanizer and detector in the same editor is actually useful, not just a marketing claim.

The one thing that could be better: the free trial word limit is a bit tight if you want to properly test it on a full JD rewrite. I got around it by breaking text into chunks during the trial, but it’s worth knowing about before you start.

Is Walter AI worth paying for? For me, yes. The value isn’t doing something I couldn’t do otherwise. It’s doing something faster and with less anxiety about whether the tone lands right. I send better messages more consistently and I don’t second-guess every word before hitting send.

If you want to try it before committing, the free trial at walterwrites.ai doesn’t need card details. That’s how I started three months ago and I’m still here.

If you’re in recruitment or HR admin and you’ve got a different workflow, I’d genuinely be curious. Hit reply and let me know what you’re doing.

Reply rates on candidate outreach had dropped. Was everything ok with my process? She was being nice about it, but the question was clearly “what’s going on.”

I’d been using ChatGPT for first drafts. Had been for a while. It shaved an hour off my day and honestly I’d stopped thinking critically about what was going out. Bad habit to fall into when you’re sending 20-30 messages a week.

I told her I’d look into it and fix it. Didn’t mention ChatGPT.

What I found when I actually looked into it was more useful than I expected. Ended up running a small test comparing raw AI drafts to humanized ones on the same role, same week. The reply rate difference was 41%. I’ll explain what I did and what tools I used, because the question of which humanize AI is best comes up constantly in the HR groups I’m in and I’ve actually got a data point now.

What’s Actually Wrong With Raw AI Outreach

It took me embarrassingly long to spot it.

The issue isn’t grammar. ChatGPT is perfectly grammatical. The issue is that the messages sound like they were written by someone who has read thousands of recruitment emails but never actually sent one. You know the type of phrase. “I wanted to reach out to explore potential alignment between your profile and an exciting opportunity.” No recruiter in Stockport or anywhere else actually says that to someone on LinkedIn.

Candidates get a lot of messages. The ones that get replies tend to sound like they came from a person. The ones that don’t tend to sound like a template. AI output, by default, sounds exactly like a template because it is one.

I only clocked this when I read my own drafts out loud. About half of what I was sending I would never say to someone sitting across from me. That’s the test, tbh. If you wouldn’t say it in a conversation, why are you putting it in an outreach message.

The Actual Test I Ran

Two weeks, one role: junior ops, standard permanent position. I pulled candidates from the same LinkedIn search, split them into two groups of 30. Same job, same salary range, same criteria.

Group A got the ChatGPT draft sent as-is. Decent prompt, asked for something direct and friendly. Went straight from the model to the send button.

Group B got the same message run through Walter Writes on Standard. I read it once, made sure it still sounded like me, then sent. Extra time: about 90 seconds per message.

Five days, tracking actual replies. Not opens, actual replies.

Group A: 8. Group B: 14.

Which Humanize AI Is Best?

I tested three before settling.

Walter Writes

The AI Humanizer one that worked for me. The rewrites didn’t just change the words, they changed how the ideas were laid out. Shorter sentences where the original was long. Broken-up phrasing where the original ran on. The tone options were useful for recruitment specifically: I used casual for candidate messages and professional for hiring manager updates. The built-in AI score after each rewrite helped too. Gave me something to check against rather than just guessing.

I still read everything before sending. That part hasn’t changed and probably shouldn’t.

QuillBot

I’d used before for other stuff. It works fine for paraphrasing documents, but for outreach it was too shallow. It changes some words but the sentence structure stays the same, and structure is what flags AI text. You end up with the same robotic cadence in different vocabulary.

StealthGPT

was the aggressive option. Some of the output was usable, but it overcorrected in weird ways. One message came back so obviously “trying to not sound like AI” that it actually sounded worse than the ChatGPT original. There’s a specific kind of awkwardness that happens when something tries too hard to be casual and overshoots it.

How to Remove AI-Generated Text From Outreach

Most people asking how to remove AI generated text end up reading articles about Grammarly, which is not the answer.

Grammar checkers fix grammar. They don’t fix the structural patterns that make AI text recognizable to a human reader. What you need is something that rewrites at the sentence level, not just the surface.

Here’s what actually works.

Take your ChatGPT draft. Paste it into a proper AI humanizer, not a paraphraser, not a grammar tool. Something that does sentence-level rewriting. Pick a tone that fits who you’re writing to. Read the output once before sending.

The Walter Writes AI humanizer handled my ChatGPT and Claude drafts without much trouble. Occasionally a sentence comes out slightly off, which a quick read catches. Under two minutes per message once you’ve done it a few times.

Does It Get Caught?

Short version: the humanized output scored mostly-human on AI readability checks every time I ran it through one.

But that’s probably not the more important question for recruitment. Nobody is running your LinkedIn outreach through GPTZero. What matters is whether the candidate reads it and thinks “this was written by a real person for me, specifically.” Raw AI drafts fail that test. Humanized ones don’t, or at least mine didn’t.

There’s also a version of this question about whether applicant tracking system (ATS) software or AI monitoring tools will flag outreach, but I haven’t tested that and I’d be making things up if I claimed to know.

What 41% More Replies Looks Like in Practice

8 vs. 14 might not sound like a huge number. In practice the difference was noticeable.

8 replies from 30 messages means a lot of silence. You send follow-ups, you wonder if they saw it, you’re not sure where anything stands. 14 replies means you’ve actually got conversations to run. Candidates in the pipeline, managers to update, a process that’s moving.

The quality difference was probably more striking than the quantity. Group A replies were almost all transactional. “What’s the rate?” “How long’s the contract?” Short, no warmth, just extracting information. Group B had actual replies. A few candidates said the message felt more personal than usual. One said it was the only outreach they’d received that week that felt like it came from an actual recruiter rather than a bot.

I keep coming back to that. Candidate experience starts before the interview. It starts with the first message they get from you, and if that message reads like a template, you’ve already started things off on the wrong foot.

The other thing worth mentioning: better reply rates mean fewer follow-up messages. If your first outreach gets more responses, you’re spending less time chasing people who’ve already decided to ignore you and more time having actual conversations. That part probably saves as much time as the AI drafting did in the first place.

Does the Free Version Do Enough?

This matters to me more than it might matter to people at bigger agencies, because I’m watching every subscription.

Walter Writes gives you 300 words free, no account needed. For short outreach messages that’s usually enough to test it on a real draft and see if it works before committing to anything. Paid plans are about $8 a month on annual billing, which is roughly what I’d spend on a Friday lunch.

I’ve tried free humanizers before this. Some of them chopped sentences in ways that introduced errors. Some just changed vocabulary without touching structure, so the output still reads like AI with different words. The free trial was enough to tell me Walter Writes wasn’t doing either of those things.

If you’re doing very high-volume outreach, the word limits on the free plan won’t go far. For the kind of selective outreach most entry-level recruiters are doing, it’s a reasonable starting point.

How I Actually Use It Now

Draft in ChatGPT. Prompt for direct and conversational language. Run it through Walter Writes on Standard or Simple depending on how much work it needs. Read it once. Send.

Four to five minutes per message, for something that used to take me 15 to 20 minutes at equivalent quality. The reply rate improvement is real and it’s held over three months, not just the initial test.

One caveat worth being honest about: humanization sorts the tone problem, not the substance problem. If your job description is vague or the brief from the hiring manager is useless, a better-sounding message still won’t land. The tool helps you sound like a person. What you say is still on you.

But if you’re sending solid messages and still getting silence, the sound of them is probably the issue. That’s the bit that’s actually fixable in under two minutes.

Happy to dig into specifics if anyone wants to compare notes. Especially if you’re working sectors where reply rates are low to begin with, would be good to know if others are seeing similar results.

Her reply was two words: “ChatGPT much?”

Not mean, not even really complaining. Just... tired-sounding. Like she’d been getting them all week, which she probably had.

I kept that message in a folder on my desktop for a while as a reminder to actually sort out my outreach process. I’m an entry-level recruiter, doing volume work — reaching out to passive candidates, following up, chasing hiring managers, keeping everything moving. The AI drafts are a necessity. There’s genuinely no way I’m writing every single message from scratch. But clearly the raw AI draft is a problem.

I spent a few weeks testing different humanizers to figure out what actually works for this specific use case. Here’s what I found.

The short version: Walter Writes is what I settled on, and it genuinely changed how my messages land. HumanizeAI.tech is a reasonable option if budget is the main concern. Everything else I tested either didn’t do enough or made things worse in a different direction.

Why Cold Outreach Is Specifically Hard to Humanize

Cold outreach has a specific quality requirement that differs from other kinds of writing. The goal isn’t to sound professional. It’s to sound like a specific person sent a specific message to a specific person. That’s a different ask.

AI drafts cover all the bases. They’re efficient. They don’t miss anything. But they also don’t have the small deviations that come from a person writing with a particular candidate in mind. The rhythm is too even. The transitions are too clean. The level of warmth is calibrated but never quite right. Basic humanizers fix the most obvious signals — they’ll remove the really stilted phrasing — but they don’t fix the underlying evenness. The message still feels assembled.

The other thing that matters for outreach specifically is the tone target. Too formal and it reads like a form letter from HR. Too casual and it doesn’t match the professional context. That’s a narrow band to hit. Most humanizers apply a general “more human” transformation that doesn’t reliably land in it.

Walter Writes

The tone setting is what won me over. Before you run the humanization, you can specify what register you want — casual, academic, journalistic, brand-safe. For recruitment outreach I almost always use casual. That’s not just pattern removal. It’s rewriting toward a specific register. The messages come out sounding like something a recruiter actually typed rather than something an HR template generated.

The structural rewriting also goes deeper than the basic paraphrasers. It changes how ideas connect, how long sentences run, where the rhythm breaks. That variety is what makes a message read like a person was behind it. I sent similar outreach to two comparable pools over a few weeks — one batch with Walter, one I’d edited manually — and the Walter batch got noticeably more replies. I wasn’t expecting the difference to be that clear, but it was.

Most of my outreach goes through LinkedIn and Gmail. The Chrome extension is useful because it lets me humanize without leaving the tab I’m on. No copying and pasting between tools. I’ve been on the free tier mostly, which gives you 300 words per session and covers individual messages fine. For longer sequences or detailed job descriptions, the Starter plan is around £80 a year. I can justify that given what it saves.

One thing worth noting: this isn’t about detection in the academic sense. Nobody’s running your LinkedIn message through Turnitin. The problem it solves is candidates recognising the feel of AI-generated text — which they increasingly do — and the response rate hit that comes with that.

HumanizeAI.tech

Humanize AI is worth knowing about, mainly because it’s cheaper at the entry tier and it handles the basics reasonably well.

The processing is quick, the interface is simple, and for a straightforward “make this ChatGPT draft sound less robotic” job it gets there. Fewer stiff transitions, less of the obviously formal opener language. If your budget is very limited and you’re doing lower-volume outreach, it’s a workable option.

Where it falls short is in how specific the tone becomes. The output reads more human, but it doesn’t consistently read like me. The warmth it adds can feel a bit applied — present but generic. Candidates who are getting a lot of outreach (which good candidates are) tend to pick up on that, even if they couldn’t explain why. It solves the obvious problem but not the underlying one.

I’d use it for situations where I just need the basic AI tells removed and don’t need the message to sound particularly like me. For most outreach where I’m trying to actually get a response from someone I want in the pipeline, I’d stick with Walter.

What I’d Actually Suggest Trying

Pull out the last ten outreach messages you sent and run them through Walter Writes on the casual setting. Read them out loud. If they sound like something you’d say to a candidate you were genuinely interested in reaching, good. If they sound like a well-structured notification, they need another pass.

Tracking response rates over a few weeks is what actually convinced me the difference was real and not just me expecting it to work because I’d paid for it. The humanized messages did better, consistently. That’s enough justification for a Starter plan subscription at the volume I’m working.

If that experience with the “ChatGPT much?” reply sounds familiar — happy to hear if others are handling this differently. Drop a note in the comments if you’ve found something that works better.

I clean up job postings that have been drafted with generic or vague wording. Afterward, I find that candidates whose qualifications match our company’s ideal profile are more likely to respond to our job posting. My goal is to replace vague, generic words with descriptive words that give potential employees a clear idea of what our company is looking for.

Using vague words in a job description is problematic. Many of these terms are used generically and therefore lack specificity. For example, many companies claim to be seeking “a dynamic and results-driven individual.” While being dynamic and results-driven may be desirable characteristics for any new hire, this term lacks specificity.

What does this mean? Is the person expected to think creatively, produce measurable results, or take initiative? The term provides no direction for potential applicants to determine whether they’d be a suitable fit for the position. It provides the reader with very little information that would allow them to determine whether they meet the expectations of the job.

Another example is the term “fast paced.” This term describes the type of environment the new hire will be entering. It fails to provide any information regarding the types of responsibilities the new hire will face. Is the pace due to a large volume of work, constant changes in procedures, or a competitive environment? The term “fast paced” should be replaced with a term that specifies what aspects of the job will require the ability to work at a fast pace.

Terms such as “we are a team player” fail to provide any valuable information. Almost every employer claims to desire individuals who are team players. However, employers typically seek different types of team players. Some want individuals who are willing to assist colleagues at random, while others prefer workers who focus primarily on completing assigned tasks.

Some employers include the expectation of strong communication skills within the list of qualifications for each job. While strong communication skills are necessary in virtually all positions, they’re particularly important in jobs involving contact with customers or clients. You should define the types of communication and levels of communication required in each position. In a sales position, communication may involve frequent phone calls to customers. In a production position, it may occur primarily via email.

Using vague terms in a job description will only serve to deter applicants. To create an effective job description, use detailed and specific terms that clearly describe the expectations of the position. This will allow potential employees to assess whether they meet the expectations of the job and make informed decisions about their applications.

“Must Be a Team Player”

I recently rewrote a job description for a recruiter and took out the phrase “We’re looking for someone who’s great to work with.” My boss wanted it back in. She explained that her team needed someone who was easygoing and worked well with others.

While I agree with her sentiment, the issue lies in how poorly “team player” screens for it. Nobody has ever written “terrible team player” on their resume. People always claim to be great team players. So it’s basically useless as a filter.

A better way to screen for it is to describe how the team works. For example:

• Do they hold daily meetings?

• Do they work independently and only come together for planning sessions?

• Are they responsible for covering each other’s workload when one team member is overwhelmed?

Describe that stuff and candidates will figure out if they’d thrive in that environment. But use a catch-all phrase like “great team player” and you’ll just get candidates who say whatever sounds good but don’t really intend to follow through.

“Competitive Salary”

Job seekers know what “competitive salary” means. It means either we don’t want to give you the salary number yet, in hopes you’ll magically guess it, or the salary is below market and we hope you won’t realize it until we offer you a contract.

While I’m sure some organizations are truly prevented from listing salaries internally, for example due to equity considerations or management policies outside of my control, I’ve noticed that smart job seekers immediately start calculating whether it’s worth spending their time interviewing us when they have zero clue whether the salary will be viable.

If possible, I always advocate for listing a salary range, even if it’s just a rough estimate, simply because it allows candidates to self-filter and avoid wasting their time. Even if you can’t convince your organization to provide a hard dollar amount, I at least encourage removing “competitive salary” and replacing it with something concrete that speaks to the value proposition, like “We offer flexible hours” or “You get access to remote work.”

“Fast Paced”

To me “fast paced” is another lazy way of saying “high volume work” or “multiple competing priorities.” Problem is, those are completely subjective experiences based on what someone else considers “fast paced.”

And even worse, if it’s not true, in other words your team isn’t actually moving that fast, you risk attracting someone who expects chaos but ends up bored stiff when they discover reality is far slower than advertised.

So again, I think describing what makes your work environment fast paced is key. For example, “This role manages 80+ candidate communications per week across 5 open roles” gives a better sense of what it means to be fast paced versus simply saying “recruitment environment is fast paced.”

“Dynamic & Results Driven Individual”

This one appears frequently and tells me next to nothing. We all want someone who’s “results driven,” but if that’s literally all I’m getting from a posting, and it’s repeated several times, it feels like the writer did the minimum amount of work required.

I’m not arguing against wanting someone results driven. Most organizations do. The argument is that telling me that tells me nothing. What do you really want this person to accomplish? Do you want someone who doesn’t need hand holding? Who crushes performance metrics? Who can handle a high volume of activity?

All of those are legitimate desires. None of them require the blanket statement “dynamic & results driven.” Remove the blanket statements and add specifics.

“Strong Communication Skills Required”

This is perhaps the most common example I see of a generic term applied broadly. Since every single role requires communication, yes, even if you’re mostly working solo, stating “strong communication skills required” doesn’t tell me what kind of communication I’ll be doing (written versus verbal), in what context (to whom, when, or how often), at what volume (how many messages or reports), or how it applies to my role.

In most cases, when recruiters see “strong communication skills required,” they interpret it as “you’ll be writing lots of emails to candidates,” rejection letters included. That’s worth specifying. “High volume written candidate communication, including rejection emails, update emails, and interview invites” tells me something real. “Strong communication skills required” tells me nothing I didn’t already assume.

Same goes for other roles: presentations, customer service calls, and departmental collaboration. Each of those has specific implications for communication requirements that can be specified explicitly. None of them require “strong communication skills.”

If I’m rewriting a job description and I encounter any generic terms listed above, my standard procedure is: can I specify what this term really means? Can I make this term more specific? If neither option works, in other words the term is entirely vague and nothing can be reasonably inferred, I remove it.

This article will provide the practical view. What is actually being used at the day-to-day level. What works. And what are the limitations. As an entry-level recruiter based in Stockport, I’ve developed my own experience with the tools referenced below. I’ve also talked to many other individuals in similar junior positions. They share my experience.

The practical stack: a basic grammar checker & ChatGPT + humanizer tool

A ChatGPT AI humanizer tool is the best option for creating first drafts of communication. There are no costs associated with ChatGPT for the basic version, it’s extremely fast, and it creates a foundation to start drafting communications. There are limitations to using ChatGPT for creating first drafts. Knowing these limitations prior to starting may help minimize frustration.

All of the raw ChatGPT generated output for HR and recruitment related topics default to a formal register. While this formal register may be great for creating documents that need to be formalized, it’s not how humans communicate. This formal register can create a sense of robotic communication, which can cause candidates to question whether they’re communicating with an individual or a computer program. In many cases, candidates respond to highlight this difference.

However, editing can resolve the formal register issue created by ChatGPT. The key factor in determining how much editing is needed depends greatly upon how vague or specific your prompt was. Vague prompts will generate generic responses, which will require a substantial amount of time and effort to revise to meet your intended standards. On the other hand, if you create a specific prompt that incorporates tone guidance and provides examples of how you intend to write, the response will likely be close to complete or even complete.

While many users of ChatGPT learn through trial and error, I learned quickly through the same method. I drafted numerous communications that were approximately 70% completed and questioned why the output didn’t seem natural until I realized that the solution wasn’t in ChatGPT, but rather in my prompts.

What do AI humanizer tools really do?

An AI humanizer tool takes AI-generated text and rewords it to eliminate characteristics that identify it as being generated by a machine. These characteristics typically include uniform sentence structures, formal phrasing constructions, and consistent rhythms that trained readers associate with AI-generated output.

When evaluating AI humanizers, there are two distinct categories to consider: shufflers and rewrite engines. Shufflers produce output that sounds different but still processed, whereas rewrite engines produce output that your editing pass can actually work with.

For recruitment writing specifically, the primary objective is to address the formal register produced by AI tools prior to completing an editing pass. Manual replacement of each formal phrase is one of the most laborious components of this process. Finding a reputable AI humanizer tool that addresses this task automatically is critical to achieving efficiencies in your workflow.

Humanizer tools function at a technical level

At a technical level, humanizer tools analyze statistical patterns within the text, including perplexity scores, burstiness, and transition frequencies, to reword them to increase variability. Better tools accomplish this at the sentence structure level rather than simply substituting synonyms. Substituting synonyms retains the same structural DNA, whereas rewriting at the structure level alters how concepts are communicated, which is the very action that decreases detection scores.

What are the actual tools used by entry-level recruiters?

As mentioned above, many recruiters at entry levels use free AI tools due to budget constraints. Although numerous free humanization tools exist and are certainly worthy of testing if budget is truly an issue, none perform uniformly well on short-form content nor do they necessarily introduce fewer issues. Be sure to evaluate performance on your actual content as opposed to demo material. If you find free tools to be sufficiently reliable for your application purposes, then that should be sufficient.

I personally use WalterWrites, a paid AI humanizer tool, and it appears to be frequently discussed among both recruitment and HR professionals in the groups I participate in relative to producing AI communication that sounds human. It continues to appear throughout our discussions due to its functional nature: it effectively eliminates formal register issues more reliably than many free alternatives and minimizes post-humanization editing time. With pricing at approximately $96 annually ($8 monthly) at the Starter level, I think it’s completely justifiable to invest in such a tool on an entry-level salary if using AI-drafted communications regularly.

Specifically regarding rejection emails and candidate update messages, which represent two primary applications of WalterWrites, the tool successfully accomplishes both tasks without generating additional problems. The built-in score comparator allows you to visualize pre-and-post-score improvements prior to sending communications. According to tests conducted by WalterWrites, their outputs decrease from over 95%+ AI detected via GPTZero, Turnitin, Originality.ai, and Copyleaks down to less than 5%.

In conjunction with the use of WalterWrites’ humanizer tool, I also use Grammarly at the free-tier level for error-checking purposes. While not a humanizer tool itself, it functions well enough for detecting errors prior to sending communications.

Practical workflow that produces good output

To date, regardless of whether I’m writing rejection emails, inviting candidates for interviews, rewriting job descriptions, or updating candidates who have waited on status messages, I’ve found that the following workflow generates reliably good output:

1. Create a specific ChatGPT prompt with tone guidance

2. Generate ChatGPT draft

3. Apply a humanization pass using WalterWrites

4. Complete an editing pass for tone and specificity

5. Review final output prior to sending

If I’m only concerned with low-stakes internal communication like contacting a hiring manager or providing updates to colleagues, a ChatGPT draft plus a rapid edit is often sufficient.

One common temptation at scale is to bypass both the humanization and editing passes altogether and simply send the AI-created draft. Each time I’ve chosen to skip both steps and send the AI-created draft under time-pressure is when I’ve subsequently needed to send a follow-up communication because something didn’t sound right.

On average, this entire workflow takes 5-7 minutes per message when developing something akin to a rejection email. Comparatively speaking to writing from scratch, that’s fast. Blank-page syndrome ceases to exist. Anxiety concerning whether you’ll adequately articulate yourself in your phrasing diminishes considerably as you’re able to move along with the process instead of second-guessing yourself.

Common questions regarding using AI humanizer tools for recruitment communications

Will candidates or colleagues know I’m using AI?

Only if you skip or rush the editing step after applying a humanization pass. The purpose of a humanization pass is to remove all of the AI patterns present in your formal output. After a proper humanization pass and subsequent editing pass, you can’t tell your output was created using an AI tool. It reads like it was authored by an individual.

Are AI tools ethically acceptable for composing candidate e-mails?

Ethics regarding the use of AI for professional communications boil down to two factors: honesty in your messaging and treating individuals appropriately with regard to their dignity and respect. An AI-assisted drafted rejection email that’s personalized, clear, and respectful is superior to a cold-form rejection email authored by a human.

Do company-wide AI usage policies impact your ability to use AI tools for candidate communications?

From my perspective, and from the perspectives of nearly all junior recruiters I’ve spoken with, they operate in gray areas since neither their employers prohibit nor endorse their use of AI tools. My general operating procedure is: use AI for drafting communications; ensure all output is thoroughly reviewed and revised prior to transmitting; never input candidate-specific data into AI tools if possible.

Has the humanizer worked on larger content items such as job postings?

Yes, it works particularly well on larger content items such as job postings, where time savings will be greatest. A 600-word job posting that requires humanization will take around 2 minutes using WalterWrites and produce an output that reads naturally. Creating a job posting from scratch or manually revising one would take significantly longer.

The cost consideration for junior recruiters

Tools cost money, and junior recruiters don’t have much left over.

My philosophy: use free tools first. Upgrade only where free versions demonstrate insubstantial performance relative to volume and quality expectations.

• ChatGPT free tier for drafting communications

• Free grammar checkers for basic error detection

• Paid humanizer only when free alternatives fail consistency-wise

What I don’t do: pay for multiple tools that serve substantially similar purposes or tools containing features I don’t anticipate using.

Keep it simple. If any tool doesn’t save time or reduce stress sufficiently, cancel it.

My test for paying any tool: would I be able to justify paying for this tool based solely on time saved or stress relieved? For WalterWrites, yes. Time-saving reductions per week across the board, alongside stress-reduction associated with sending potentially perceived “robotic” communications, are tangible realities.

What nobody told you about using AI tools in recruitment

Efficiency gains are achieved by speeding up your editing process, not by eliminating it.

You’re not delegating decision-making authority regarding what you say or how you say it. You’re eliminating the blank-page problem and formal-register problem.

Actual communication decision-making remains entirely yours.

Entry-level recruiter. Limited budget. Using AI constantly to keep up with communication volume but frustrated by how obviously robotic the output sounded. Looking for something that would close the gap without costing more than I could reasonably justify on an entry-level salary in Stockport.

That’s the lens this review is written from. Not a marketing team with a budget and a strategic content operation. Just someone doing admin recruitment work who needed AI drafts to sound like a person wrote them.

WalterWrites is an AI humanizer and detector that rewrites AI-generated text to remove the patterns that make it identifiable as machine-generated, then gives you a score showing how human the output reads. After a few weeks of using it in my actual recruitment workflow, here’s what I found.

For HR and recruitment writing specifically, WalterWrites does what it claims. The formal register problem that plagues raw ChatGPT output gets addressed consistently enough that my editing time per message dropped noticeably. It’s not a perfect tool and it won’t write the email for you, but as a middle step between AI draft and final human edit, it delivers.

What I Was Trying to Fix

Raw ChatGPT output for recruitment communication has a consistent problem. It’s too formal. Technically correct, tonally wrong.

“We regret to inform you that following careful consideration of your application, we have decided not to progress your candidacy at this stage. We wish you every success in your future endeavours.”

Nobody writes like that. Nobody wants to receive that. When you’re writing rejection emails, candidate update messages, and job description rewrites all day, the AI formal register adds up to a significant amount of editing time.

I was manually fixing the same constructions in every draft. Removing “we regret to inform you.” Cutting “following careful consideration.” Replacing “at this juncture” with literally any normal human phrase. It was taking long enough that the efficiency gain from AI drafting was mostly being eaten by the editing time.

I needed a middle step that handled the formal register problem before I got to editing.

What I Tested Before Committing

I tried free humanization tools first. Budget constraint being what it is, free seemed like the obvious starting point.

The results were inconsistent. One tool produced output that was different from the original but introduced its own awkward phrasing that still needed editing. Another improved things slightly but not enough to change my workflow. Neither was reliable enough to become a standard step.

I also tested Undetectable.ai. The results were better than the free tools, but the pricing and interface weren’t working for me at the volume I was running. Worth saying that clearly — different tools work for different workflows.

WalterWrites came up in a couple of Reddit threads when I was searching for humanizers specifically for professional email writing rather than academic content. I tested the trial on my actual content rather than the demo text — rejection emails, job description rewrites, candidate update messages.

What the Trial Actually Showed

The first thing I noticed was the editing time difference.

After a WalterWrites humanization pass, the formal constructions were consistently gone. The output wasn’t perfect — it still needed an editing pass for tone and specificity. But I wasn’t spending that editing pass fixing systematic formal register problems. I was spending it on judgment calls: does this specific message need more warmth, is this phrasing right for this candidate’s situation.

That’s a smaller task. Not because the total editing effort was much lower in absolute terms, but because fixing formal register problems manually is repetitive and draining in a way that making judgment calls isn’t.

For rejection emails specifically, the humanized output required noticeably less work to get to the standard I needed. A five-minute editing pass rather than a ten-minute one isn’t dramatic, but at the volume I’m working at across a week, it adds up.

The built-in AI detection score is useful too. After humanizing, I can see immediately whether the output is likely to read as AI-generated. WalterWrites shows you the before and after score, so you can see the actual difference. On rejection emails I drafted, I was typically going from 90%+ AI down to under 10% after humanization and a short edit. That’s within the range that reads as human-written.

What WalterWrites Actually Does

It’s worth understanding what separates a tool like WalterWrites from just asking ChatGPT to make something “sound more human.”

ChatGPT-to-ChatGPT editing tends to preserve the underlying patterns that make text detectable. The model rewrites but keeps the same structural DNA — uniform sentence length, the same transition rhythms, the same phrasing cadences.

WalterWrites rewrites at the structure level, not just the surface. It changes how ideas are expressed, not just the words used to express them. The result is text with more natural variation in sentence length and rhythm — the kind of burstiness that human writing has and AI writing consistently lacks.

It also has three rewrite strength settings: Simple, Standard, and Enhanced. Standard is sufficient for most of my use cases. For something that came out particularly robotic and formal, Enhanced gives you a more aggressive transformation. Having the option to adjust that depending on the starting quality of the draft is genuinely useful.

The Cost Question

I’ll be honest: the pricing required some thought on an entry-level salary.

WalterWrites’ Starter plan is $96 per year — around $8 per month. On my salary, that’s a real monthly cost. I had to actually think about it rather than just signing up.

The way I justified it: the time saved in editing across a week was real. The anxiety reduction of not having to second-guess every AI draft quite as much was also real, even if I couldn’t quantify it. Together those felt worth the subscription cost.

I’d be more hesitant recommending it to someone producing low volumes of content where the editing time saved is minimal. At that level, careful manual editing of AI drafts might be sufficient. At the volume I’m producing, the subscription is worth it.

Limitations Worth Knowing

It doesn’t write the message for you. The editing pass is still necessary and the final message still needs your actual judgment about tone and content.

For very short messages — one or two sentences — it sometimes doesn’t make much difference. The simpler the original draft, the less there is to improve.

The humanization can occasionally flatten nuance. If your original AI draft had a specific tone or phrasing you wanted to keep, you sometimes have to restore it after the humanization pass. Not a common problem, but worth doing a read-through for it.

And like any tool in this category, it’s a step in a workflow, not a replacement for the workflow. It makes the editing stage more efficient, not unnecessary.

The Short Version

Does what it claims, reliably enough that it’s become a standard step in my workflow. The formal register problem that makes raw ChatGPT output unsuitable for candidate communication gets handled consistently. The cost is justifiable if you’re producing regular volumes of AI-assisted professional communication on a salary that requires actual cost justification.

Test it on your actual content during the trial — rejection emails, job descriptions, candidate update messages, whatever you actually send. The 300-word free trial at walterwrites.ai/ai-humanizer is enough to run a real test without committing to anything.

Happy to answer questions from others in similar junior or entry-level positions. Budget-conscious tool decisions are worth talking about honestly.

A significant chunk of my working week looks like this. I’m an entry-level recruiter, and managing the gap between candidates who need information and hiring managers who are busy and not prioritising recruitment communication is one of the main things I actually do.

I don’t have a solution that eliminates this. But I’ve found approaches that make it less draining.

The practical way to handle hiring manager silence is to separate what you can control from what you can’t. You can’t force a faster response. What you can do is make replying easy, template your candidate holding messages so they don’t eat mental energy each time, and know when to escalate rather than just keep chasing.

What You Can and Can’t Actually Control

Hiring managers not replying promptly is not a problem I can fix. I can chase. I can make it easy for them to respond. I can flag urgency when it exists. I can’t make them prioritise feedback.

Early in the role I spent a lot of mental energy being frustrated about this. It didn’t help anything and it made the job harder than it needed to be.

The acceptance isn’t passive. I still chase, I still follow up, I still make the case for timely communication when I have the opportunity. But carrying frustration about something outside my control as a constant background stress is just an energy tax with no return.

What I can control is how organised I am when chasing. When I have a clear record of when the interview happened, when I last chased, and what the candidate’s current expectation is, I’m less stressed because I know exactly what’s outstanding. The chaos happens when I’m tracking it in my head rather than somewhere I can actually look at.

I use a simple spreadsheet. Role, candidate name, interview date, last chase date, candidate’s last contact date, current status. Not sophisticated. But reviewing it at the start of the day tells me exactly where the pressure is.

Chase Messages That Make It Easy to Reply

The default chase email is something like: “Just checking if you have any feedback on the recent interviews?”

The problem with this is it requires the hiring manager to compose a response. They have to think about what their feedback was, formulate it, and write it out. That’s friction. It makes it easy to deprioritise.

I’ve had better response rates with messages that make replying simple. Something like: “Just checking in on the interviews from last week — happy to draft the feedback summary if you can give me a couple of bullet points on each candidate. Let me know if that works.”

That’s a smaller ask. They don’t need to write a structured response. They just need to send a few notes. The smaller I can make the effort required to respond, the more likely I am to get something back.

A loose deadline sometimes helps too. “No rush, but the candidates are following up — if you can get me something by Thursday I can update them before the weekend.” An actual date makes it more real than an open-ended request that can drift indefinitely.

I’ve also started being more explicit about the candidate impact when chasing. Not guilt-tripping, but factual: “The candidate who interviewed on Monday has reached out twice — just so you know the timeline from their side.” Some hiring managers genuinely don’t think about what the silence feels like from the candidate’s perspective. A brief note about it sometimes moves things.

Templating Holding Messages for Candidates

I have a few versions of holding messages that I rotate. Different enough that I’m not sending the exact same email three times to the same person, similar enough that I’m not writing from scratch each time.

The core message is always the same: we’re still in process, no decision yet, I’ll be in touch when I have something. The variation is in framing and phrasing rather than substance.

Version one: “Just wanted to keep you updated — we’re still in process and I hope to have something to share with you by [rough timeline]. Thanks for your patience.”

Version two: “Wanted to drop you a quick note — still waiting on the decision from the team, but you’re still very much in the process. I’ll be in touch as soon as I have an update.”

Version three: “Just checking in on my end — no news yet, but I wanted to make sure you knew it hadn’t been forgotten. Should have more clarity by [date] and I’ll be in touch then.”

Simple, brief, honest. None of them give false information or pretend things are moving faster than they are. They just confirm the candidate hasn’t been forgotten.

This is also somewhere AI drafts are actually useful, ngl. Getting a quick first version of a holding message that I can personalise and send takes much less time than writing from scratch. The editing is minimal because the stakes are low and the content is simple. The main thing I’m checking is that it doesn’t sound like a form letter.

When the Standard Chase Isn’t Working

Most hiring managers respond eventually. Some don’t.

If two chases haven’t produced anything in ten days, I’ll try a different channel. Slack or Teams often gets a faster response than email. Switching medium is sometimes all it takes.

If that still doesn’t work, I’ll try reframing the request. Instead of asking for feedback, I’ll ask something simpler: “Did you want to progress any of the candidates, or should I close out this round?” A yes/no question is easier to answer than an open-ended feedback request.

For very busy hiring managers, offering to handle the whole thing sometimes works. “If you can confirm whether you want to progress [name] or not, I’ll handle the communication to all candidates — you don’t need to write anything.” Taking the task fully off their plate reduces the barrier to a single reply.

When to Escalate

I try not to escalate small things. I’m aware I’m junior enough that constant escalation is probably not a great look. Most of the time, persistent chasing eventually gets results.

But there are situations where escalating is the right call.

If a candidate has been waiting more than two weeks for post-interview feedback with no communication, that’s a genuine candidate experience failure that could generate a Glassdoor review or a complaint. Flagging it more directly to whoever manages the hiring manager is reasonable at that point — not aggressively, just factually: “Just flagging that the candidate from the 2nd is still waiting on feedback — it’s been two weeks and I don’t want this to become a complaint.”

If the delay is actively affecting the process — a strong candidate is likely to accept another offer if we don’t get back to them — that’s worth escalating too. Framing it in terms of risk rather than frustration tends to land better.

The pattern I follow: one more chase, then flag to an appropriate next step, then I’ve done what I can do.

The Mental Load of Being in the Middle

The part nobody really prepares you for when you start in recruitment is the weight of being the person who holds information that other people are waiting for.

Candidates contact me because I’m the point of contact. They want an update. I understand why. But when I don’t have one, there’s nothing I can do except acknowledge their frustration and send another holding message. Doing that repeatedly, especially for the same candidate over multiple weeks, is genuinely draining.

What’s helped me is being honest with candidates about the situation earlier rather than later. Not in detail — I’m not going to say the hiring manager has gone quiet — but something like: “The decision is sitting with the hiring team at the moment and timelines can shift. I’ll reach out as soon as I have anything concrete.” That sets a realistic expectation without placing blame or giving false reassurance.

I’ve also had to stop internalising candidate frustration as my failure. When someone emails me angrily about a delay, that’s frustrating. But the delay is usually not something I caused or can fix. I can respond with care and keep them updated. That’s the job.

That distinction took me a while to actually internalise rather than just understand intellectually.

The One Thing That Helps Most

If I had to pick a single practice that makes the most difference, it’s setting expectations with candidates before they need to chase.

If I know from experience that post-interview feedback usually takes five to seven days with a particular hiring manager, I tell the candidate that after their interview. “The hiring team usually takes about a week to get back to me — I’ll be in touch as soon as I have an update, but no rush from your side until then.”

Most of the anxious follow-up emails I get come from uncertainty. Candidates don’t know if no news is normal or a bad sign. Giving them a realistic timeline up front — even a rough one — reduces that anxiety and reduces the follow-up volume significantly.

It doesn’t solve the hiring manager communication problem. But it takes a lot of pressure out of managing it.

How do others handle the hiring manager communication problem? I don’t think there’s a clean fix, but curious what approaches help.

Then I read it. And it was immediately obvious nobody wrote it. It was a template. Someone had pressed send the same way you’d close a browser tab. I was a task that got completed.

I know what that feels like. Now I’m on the other side of it, writing rejection emails for a living as an entry-level recruiter. I think about those experiences every time I draft one.

Most recruitment teams use AI drafts now. ChatGPT is everywhere. The problem isn’t using AI. It’s that raw AI drafts sound like the template problem I just described, just with different words. Making that output undetectable isn’t about cheating anything. It’s about making sure the person on the receiving end doesn’t feel like they were processed.

The way to make AI text undetectable in recruitment emails is to run a humanization pass before editing. Raw ChatGPT output defaults to a formal register that’s technically correct but tonally wrong. A humanization tool strips the surface-level AI patterns, then you edit for tone, specificity, and voice. That three-step workflow — draft, humanize, edit — is what closes the gap between machine output and something that reads like a person sent it.

Why Rejection Emails Are Almost Always the Same

Volume is the honest answer. When you’re managing five open roles with thirty applicants at each stage, the idea of writing something specific for every rejection is unrealistic. Templates aren’t malicious. They’re a practical response to a real constraint.

The problem is that templates have become the default rather than the minimum acceptable standard. Companies that could easily do slightly better don’t bother, because nobody’s measuring candidate experience at the rejection stage. The cost of a cold email isn’t immediately visible.

But the cost is real. Candidates talk. Glassdoor exists. The people you reject today might be customers or referrers later. The way you treat people when you don’t need anything from them says something that polished employer branding can’t fix.

I’ve worked with hiring managers who didn’t realise how cold the rejection emails going out under their company name actually were. They never saw them. They were busy. The default template was the one I inherited when I started, and nobody told me to update it.

I updated it anyway. It took longer than it should have, mostly because I was figuring it out alone.

The Problem With Raw AI Drafts

ChatGPT is useful for getting something on the page quickly. The issue is what it defaults to when generating professional communication.

Raw ChatGPT output for HR contexts sounds like this: “We regret to inform you that following careful consideration of your application, we have decided not to progress your candidacy at this stage. We appreciate the time you invested in this process and wish you every success in your future endeavours.”

Nobody talks like that. Nobody wants to receive that. It’s technically correct and completely impersonal. Any candidate who’s been through a few application cycles will recognise the pattern immediately.

The formal register is the main problem. ChatGPT defaults to it because it’s trained on a lot of professional writing, and professional writing tends to be formal. For recruitment communication specifically, that formality is what makes AI-generated messages feel machine-generated rather than written.

There’s also a rhythm issue. AI-generated text tends to have consistent sentence length and structure that experienced readers notice even if they can’t name it. The same transition patterns. The same paragraph shapes. It doesn’t read like a person because it doesn’t vary the way human writing does.

I’ve had candidates reply to emails of mine asking if they were automated. That’s the failure state. It wasn’t a tone disaster. It was a raw AI draft I sent under time pressure without enough editing. That happened once. I changed my workflow after.

How to Make AI Text Undetectable: The Three-Step Workflow

My workflow isn’t complicated, but it has three distinct steps and each one matters.

Step one: prompt with specificity. The more specific your ChatGPT prompt, the less editing you’ll need later. “Write a rejection email for a candidate who made it to final interview but we went with someone who had more direct experience in X” produces something far more usable than “write a rejection email.” Generic prompts produce generic output. Specific prompts give you something to work from.

Add tone guidance too. “Friendly but brief, avoid corporate phrases, write like a human recruiter” makes a real difference to the starting draft. Not foolproof, but it reduces your editing load.

Step two: humanization pass. This is where a tool like WalterWrites comes in. After I have the ChatGPT draft, I run it through the humanizer to remove the formal register before editing. WalterWrites rewrites at the sentence structure level — it doesn’t just swap synonyms, it changes how ideas are expressed. The stiff constructions get replaced with something that reads naturally.

Before the humanization pass, a typical ChatGPT rejection draft scores around 90-95% AI on detection tools. Post-humanization, it’s generally well below that threshold. That’s not the goal in itself. The goal is that the output sounds like a person wrote it, which is what those scores are measuring.

I’ve tried free alternatives first. The results were inconsistent enough that I kept coming back to WalterWrites for anything I actually care about getting right. Free tools either introduce their own awkward phrasing or don’t handle the formal register thoroughly. For rejection emails specifically, inconsistent results aren’t acceptable.

Step three: editing pass. Humanization handles the surface patterns. The editing pass is where I add specificity and get the tone right.

For a rejection at the early screening stage, brief is fine. The candidate didn’t invest much time, and a short clear message is honest. For someone who made it to final interview, more care is warranted. What specifically was the deciding factor? Even a sentence or two of genuine specificity changes how a rejection lands.

I also do a final read where I ask: would I be okay receiving this? Not happy — nobody’s happy about a rejection — but okay. Did a person write this. Does it treat me like I exist. If yes, it goes.

What Good Rejection Emails Actually Look Like

The gap between a cold template and something that feels human is smaller than most teams think.

Specificity over generality. “We’ve decided to move forward with other candidates” is technically true but meaningless. “We had strong applications from candidates with direct experience in X, which was central to this role” gives the person something real. It’s still a no. It just sounds like a no from someone who noticed what they applied for.

Brief is fine. Absent is not. The candidates who are most frustrated are almost always the ones who heard nothing, not the ones who got a brief, clear rejection. A two-sentence email beats weeks of silence every time.

Skip the hollow phrases. “We were impressed by your background” in a rejection from a company that clearly didn’t progress you reads as noise. Either mean it specifically or leave it out.

One thing I’ve started doing is varying the subject line. “Update on your application” is a dread trigger. “A decision on your [role name] application” or just the role name performs better. Small thing. Candidates notice.

The Follow-Up That Costs Nothing

If a candidate interviewed and you haven’t heard from the hiring manager in more than a week, send a holding message. “Just checking in — no decision yet, but I’ll be in touch as soon as I have an update.” Two minutes. Prevents a lot of frustration.

Most candidates don’t expect miracles from recruitment processes. They’ve applied enough times to know how it works. What they’re asking for is basic acknowledgement that they exist in the process.

I had a candidate email me three times in ten days waiting for post-interview feedback. The delay wasn’t my fault — the hiring manager had gone quiet and I was chasing. But the candidate didn’t know that. From their side, they’d interviewed and then heard nothing.

A single holding message on day four would have changed that entire experience. It took me too long to learn that proactive communication during delays isn’t extra effort — it’s damage control.

Common Questions About AI in Recruitment Emails

Does using an AI humanizer make rejection emails pass detection tools? Depends on the tool and how much editing you do after. A good humanizer handles the structural patterns that detection tools flag. WalterWrites’ tested results show post-humanization scores dropping from 95%+ AI down to under 5% on GPTZero and Turnitin. The editing pass matters too. A humanized draft you don’t edit still reads as processed. Combining both is what produces consistently good results.

What’s the difference between using a humanizer and just editing the AI draft yourself? Time, mainly. Editing a raw AI draft from scratch means finding and fixing every formal construction manually — “we regret to inform you,” “following careful consideration,” “we wish you continued success.” That takes longer than editing a humanized draft that already had those patterns removed. The humanizer handles the systematic issues; you handle tone, specificity, and anything that needs genuine human judgment.

How specific should rejection emails be at different stages? At the CV screening stage, brief is fine — confirm the decision and thank them for applying. At the interview stage, add one or two sentences about what factored into the decision. At the final stage, more detail is warranted. The more time someone invested, the more care the rejection deserves.

What Nobody Tells You About Using AI for Candidate Emails

The efficiency gain is real. At the volume I’m handling, starting from a blank screen for every candidate message would eat my day.

But the efficiency doesn’t come from removing the editing stage. It comes from making it faster. You’re not outsourcing the judgment about what to say or how to say it. You’re outsourcing the blank page problem and the formal register problem. The actual communication decisions are still yours.

I still check every message before it goes out. I still think about whether this particular candidate deserves more than the standard three sentences based on how far they got. I still rewrite sections that feel off.

AI generates the starting draft. Humanization handles the AI patterns. Editing handles tone and specificity. That three-step process is what makes the output actually useful.

tbh, the anxiety about whether something looks AI-generated mostly settles once you trust the workflow. Not completely — I still do a final read on anything important. But the constant second-guessing calms down once you have a reliable process.

What’s your approach to rejection communication? This always gets discussed at the strategic level but rarely at the “here’s how to actually write the email” level. Happy to answer in the comments.

From where I sit, the gap between that conversation and what actually happens day to day is pretty wide.

I’m not being cynical. The senior perspective is probably useful at the level it’s operating at. I just think the junior recruiter experience is different enough that it deserves its own honest description. And I haven’t seen many people write it.

Most of the job is absorbing other people’s problems

Junior recruitment roles are, in large part, communication relay. You’re in the middle between candidates who want information and hiring managers who don’t prioritise giving it to you.

Candidates email asking for updates. You don’t have any because the hiring manager hasn’t responded to your last three messages. You write back something polite that doesn’t say that directly. The candidate emails again two days later. You write another polite reply.

You’re not responsible for the delay. You can’t fix it. But you’re the person the candidate sees, so you’re the person their frustration lands on.

This is the part nobody describes in the job listing. The emotional labour of being a buffer between a broken process and the people affected by it is significant, and it accumulates.

Tbh I’ve gotten better at not taking it personally. But in the early months it felt like being blamed for things I had no control over was just a constant feature of the job.

The writing is a bigger part of it than you expect

I knew I’d be communicating with candidates. I didn’t fully appreciate how much of the job would involve the specific challenge of getting the tone right in written communication.

Every message carries risk. Too formal and it sounds cold. Too casual and it seems unprofessional. Too brief and it seems dismissive. Too long and nobody reads it.

There’s no training for this that I received. You figure it out by sending things and waiting to see if anyone complains. If they don’t, you assume the tone was okay. If they do, you adjust.

I use AI tools for drafting a lot of my communication. Just checking in messages to hiring managers. Rejection emails to candidates. Interview confirmation emails. Job description rewrites. It saves time and it reduces the anxiety of staring at a blank screen trying to find the right words.

The catch is that raw AI output in recruitment contexts sounds like HR documentation. Very correct, very formal, not very human. I run most things through a humanization tool before I edit them properly, which adds a step but means the editing pass is about improving something rather than rewriting it from scratch.

I’m aware some people would view this as somehow dishonest. I don’t, particularly. The work is still getting done. The communication still reflects judgment about what to say — I’m just not generating the first draft from nothing every time.

Career progression is murkier than it looks from outside

When you’re starting out in recruitment, the path forward isn’t especially clear.

There are mid-level roles. There are senior roles. But what exactly gets you from here to there isn’t well defined. Is it time served? Performance metrics? The right relationships? Luck?

Ngl I’m not sure. I’m trying to be reliably good at the tasks I’m given, build enough of a reputation that people trust me with more responsibility, and avoid making mistakes that set things back. That feels like the right approach. Whether it’s sufficient remains to be seen.

The uncertainty about progression is one of the more quietly stressful parts of the job. Not dramatic stress — just a background awareness that competence alone might not be enough and I’m not always sure what else I need to be doing.

The financial reality of entry-level roles

Entry-level recruitment salaries in the UK are not comfortable. Rent gets paid. Bills get covered. There isn’t much left after.

This shapes a lot of decisions that wouldn’t look like budget decisions from the outside. Which tools I pay for personally. Whether a professional development course is worth the cost. Whether I can afford to change jobs if something better comes up (moving costs, potential gap in income, probation periods at new companies).

I’m not looking for sympathy here — it is what it is, and my situation is similar to a lot of people my age in similar roles. I just think the financial reality of junior positions is relevant context that the senior perspective on recruitment tends to skip over.

What keeps it manageable

The parts of the job that feel good are genuine.

When a candidate emails back to say thank you for being the one recruiter who actually kept them updated throughout a long process, that lands. When a job description I rewrote gets notably better applications than the last version. When a hiring manager starts sending me cleaner briefs because they’ve learned what I actually need from them.

The job isn’t glamorous. It’s mostly communication admin under moderate time pressure with limited authority to fix the systemic problems that create most of the friction.

But it’s real work that has some real impact on people’s experiences. Doing it well is worth doing, even from the bottom of the org chart.

The problem is that raw AI output in recruitment contexts has a specific failure mode. It sounds like HR wrote it. Not a person, HR. The capital letters version. Formal, careful, slightly cold, full of phrases that technically communicate something but don’t sound like a human being said them.

“We regret to inform you that following careful consideration of your application...” Nobody talks like that. Nobody wants to receive a message that talks like that.

What makes recruitment AI output specifically bad

It’s not just the general AI problem of generic phrasing. It’s that AI trained on professional communication defaults to a formal register that recruitment emails don’t actually benefit from.

Recruitment communication works best when it sounds like a real person wrote it. Not casual, professional is fine but human. Like someone who has actually read your application and is taking a moment to respond to you specifically.

AI defaults to the opposite of this. It produces something grammatically correct and tonally neutral that could have been sent to anyone, by anyone, about anything.

The specific patterns I notice most: sentences that start with “I hope this message finds you well” or similar openers nobody actually uses. Phrases like “at this juncture” or “in due course.” Sign-offs that are three sentences when one would do. An overall register that feels like it was written by a committee.

The fix that actually works

ngl I’ve tried a few approaches to this and the one that works consistently is two steps rather than one.

First, run the AI draft through a humanization tool. This addresses the surface patterns, the formal constructions, the stilted phrasing, the rhythm that signals machine-generated text. I use WalterWrites for this because I found it reduced the amount of editing I needed to do afterwards compared to other tools I tried. The output isn’t perfect, but it’s starting from a better place.

Second, do a proper editing pass yourself. This is where you add the actual human element, adjusting the tone to match how you’d actually say something, making the message feel specific to this person rather than generic, cutting anything that still sounds like it came from a template.

The combination of those two steps is what closes the gap between AI efficiency and messages that don’t make candidates feel like they’re corresponding with a system.

What good recruitment communication actually sounds like

A rejection email that works isn’t lengthy. It’s clear, it’s brief, and it sounds like a person decided to write it.

Something like: “Hi [name], thanks for your time on the interview, I wanted to let you know we’ve decided to move forward with another candidate whose background was a closer match for this particular role. We’ll keep your details on file and I’ll be in touch if something suitable comes up. Good luck with your search.”

That’s it. No “regret to inform,” no “following careful consideration,” no “at this time.” Just a clear message from a person to a person.

AI can get you most of the way to this. The humanization step gets you closer. The final editing pass is what makes it sound like you actually wrote it.

The part I always check before sending

One thing I do before any AI-assisted message goes out: I read it out loud.

If any part of it sounds like something I’d never actually say to someone, it gets changed. That test catches the “I hope this message finds you well” openers and the “at this juncture” constructions faster than reading on a screen.

Happy to tweak if needed is basically my default mental state with these drafts, they’re always starting points, never finished products. The tone disasters I’ve managed to avoid have mostly been because I checked rather than just sending.

Hiring managers send across something they’ve written quickly, or pulled from an old posting, or assembled from a template that hasn’t been updated in years. My job is to make it sound like a real opportunity someone might actually want to apply for.

After doing this for a while, the patterns that put candidates off are pretty consistent.

Requirements lists that read like a wish list

“5+ years experience in X. Proven track record in Y. Strong background in Z. Must be comfortable with A, B, and C.”

Some of these are genuine requirements. Some of them are aspirational. Some of them are things the hiring manager thought would be nice but wouldn’t actually filter anyone out. Candidates can’t tell which is which.

The result is that qualified people talk themselves out of applying because they don’t tick every box, while the list does nothing useful to filter out people who genuinely aren’t suitable.

A shorter, more honest requirements list attracts better applications. Every time. The hiring managers are usually the ones who resist this change, tbh.

Salary listed as “competitive”

Candidates have seen this enough times that it’s basically a red flag now.

“Competitive” often means “we don’t want to commit to a number before we know what you’ll accept.” Candidates know this. It wastes everyone’s time when someone gets to the offer stage and the number is nowhere near what they needed.

I understand there are reasons companies don’t publish ranges internal equity concerns, negotiation flexibility, management decisions I have no input on. But from the candidate side, vague salary information creates suspicion before the process has even started.

Jargon that means nothing

“Fast-paced environment.” “Dynamic team.” “Self-starter.” “Results-driven individual.”

These phrases appear in so many job descriptions that they’ve stopped communicating anything. Candidates read them as filler. They’re probably right.

The job descriptions that actually convert, that get good applications from people genuinely suited to the role are the ones that say specific things. What does the team actually work on? What does success in this role look like after six months? What kind of person genuinely thrives here?

Specific language requires more effort to write. It’s also more useful to everyone involved.

What I try to do when I clean them up

Remove requirements that aren’t genuine requirements. Cut jargon that doesn’t add information. Make the role description specific enough that someone can imagine doing the job.

Let me know if that works is basically my mental test, would a real person read this and feel like they understand what they’re applying for?

If the answer is no, it needs more work.

The job description is the first communication a candidate has with the company. Getting it right doesn’t take long. The cost of getting it wrong is applications from people who aren’t right for the role and no applications from people who are.

What job description problems do you see most often? I’m curious whether others are dealing with the same hiring manager habits.

First, Actual Applicants Really Need to Know What’s Going On

They’re very forgiving of delays, changes, and even negative updates. But they can’t stand being ignored. They may delay applying again if they hear absolutely nothing for three weeks. If they do finally find out they interviewed and then wait an additional week or longer without hearing back, they may begin to wonder if the recruiter has lost interest in them. Even polite follow-up inquiries are met with complete silence.

More often than not, the applicant isn’t frustrated by the final result. They become frustrated simply because they feel like they don’t matter to the recruiter. I’m currently experiencing this from both sides. I realize that delays occur because hiring managers fail to communicate back to me. From the applicant’s point of view, though, they don’t know that. All they see is that they’ve invested their time and effort into the application and received no further communication from the company.

A simple note advising the applicant that we’re “still in process” and that “no decision has been made” significantly changes the applicant experience. For most applicants, sending a short update like this takes only a couple of minutes. I’ve had numerous applicants respond thanking me for updating them.

Second, Applicants Want to Be Treated Like a Human Being, Not a Piece of Data

This seems self-evident, but in the day-to-day activities of managing many applications simultaneously while under extreme pressure, it’s easy to lose sight of treating each applicant as an individual. When you manage dozens of candidates per open position at various stages of the application process, each step becomes a checklist activity: send rejection email, request candidate’s next steps regarding feedback, schedule interview, update candidate tracking tool. While candidates may recognize the tasks involved in the process, they generally feel like their humanity is absent.

Candidates typically don’t consciously recognize that they’re treated differently based on how much time the recruiter spent thinking about them versus writing the email. They do feel a distinction, though, between a message created specifically for them versus one that was automatically generated.

In my experience, using AI tools to assist with creating messages to applicants is helpful, provided that I edit the AI-generated message before sending it. Raw AI-created content tends to resemble an over-automated version of poorly worded human-generated content. While using AI to generate initial drafts allows me more time to review and revise my responses, I do take care to spend additional time personalizing my drafts before sending them.

While it may take an additional few minutes to personalize a rejection message from a robot into one written from a person, it does make a difference. I’ve received several replies from candidates simply thanking me for taking the time to inform them personally. To me, that should be expected behavior from companies.

Third, Applicants Want Truthful Information About the Available Role

Recruitment job postings frequently include vague job responsibilities, exaggerated expectations, an ambiguous salary range described as “market competitive,” and unrealistic lists of “requirements” that no single human being possesses.

Applicants have learned enough about this process that they typically read between the lines before finishing the job posting. They attempt to determine what type of actual work is required, what the true company culture is, whether the stated “requirements” are legitimate or aspirational, and if the offered salary will offset the commute.

When job postings accurately describe what a candidate needs to succeed in the role, far better qualified candidates apply. Vague postings with exaggerated opportunities attract individuals who are drawn in by exaggeration and ultimately disappointed when reality fails to meet their expectations.

I spend considerable time modifying job postings because hiring managers load them with too much jargon or require skills and attributes beyond what they truly expect from successful candidates. Convincing hiring managers to remove that language is occasionally difficult.

Fourth, Applicants Want a Meaningful Explanation During Rejection

Most recruiters use a canned rejection message to let applicants know they didn’t receive consideration. These generic statements include phrases like “We appreciate your interest...” or “Thank you for applying.... We look forward to seeing you in another capacity.”

These types of messages aren’t malicious, but they lack any tangible substance or purpose for applicants. Applicants don’t invest their time and energy into an application merely to receive a generic message in return.

I understand why generic rejection messages are used: sheer volume and fear of legal liability. Teams tend to err on the side of caution, and generic rejections allow recruiters to move on to processing other resumes faster.

Providing even minor details about an applicant’s qualifications relative to other applicants, though, can greatly improve their perception of the recruiting process. Giving applicants examples of why they didn’t receive consideration can significantly enhance their perception of the company as an employer. I strive to treat each candidate with respect throughout the entire recruitment process.

Finally, Applicants Want to Be Communicated With Following Their Interview

Interviews require significant time investments for preparation and emotional energy. Leaving an applicant hanging after an interview is extremely discourteous and disrespectful. While I sometimes face this situation due to waiting on a hiring manager who hasn’t responded, I consider it unacceptable behavior.

The quality of treatment of candidates directly correlates with perceptions of employers as organizations. People tell others about their poor experiences with recruiting processes. Sometimes these complaints are publicized online and can negatively affect a company’s image as an employer.

While I can influence certain aspects of this process as a recruiter, I’m unable to address systemic issues from where I’m sitting.