If you used AI to draft your resume or cover letter, the worry is that a recruiter will spot it and reject you. That's the smaller risk. Hiring managers say they can spot AI-written applications, but in blind tests recruiters did about as well as guessing. The bigger risk is that your application sounds like every other one in the pile.
Can employers tell if a resume is AI generated?
Not reliably. In Tilburg University's blind test, recruiters did no better than chance at telling AI-assisted cover letters from ones people wrote themselves.
These weren't amateurs. The letters were evaluated by recruiters from major Dutch companies: Philips, PwC, Rabobank, and VodafoneZiggo. Compare that with the confidence in Resume Genius's 2026 survey of 1,000 US hiring managers:
- 80% said they can identify AI-generated resumes.
- 77% said many of the resumes they see look AI-written.
- Only 4% said they don't notice signs of AI use.
Pin's AI-generated resumes report also cites a TopResume result as a test: 33.5% of hiring managers picked the AI-written resume out of four, where random guessing gets 25%. That's 8.5 points above chance. A different source describes the same 33.5% as the share who said they can spot AI in under twenty seconds, so treat that figure carefully. On either reading, it's nowhere near 80%.
What "detection" actually measures
The biggest giveaway hiring managers named was unnatural phrasing or tone, at 51% in the Resume Genius data. That tells you how the detection works. It's a reaction to tone, not forensics. Plenty of people write stiff, formal prose with no AI involved, and an edited AI draft can read like normal speech. So the "detector" goes off when writing sounds generic, whoever wrote it.
Do recruiters reject AI-written resumes?
A minority do, and it's far smaller than the number going around. TopResume surveyed 600 hiring managers through Pollfish on May 15 and 16, 2025, and 19.6% said they would reject a candidate with an AI-generated resume or cover letter.
You'll also see "80% of hiring managers would reject an AI-generated resume outright," which Gem's blog repeats. It looks like a mix-up with Resume Genius's "80% can spot it." The rejection figure with a source behind it is 19.6%.
Put those two findings together. About one in five hiring managers would reject what they believe is AI-written, and what they believe is close to a guess. So the rejection doesn't land on AI use. It lands on applications that read like boilerplate. On the candidate side, Jeremy Schifeling's Career Pulse survey found only 4% of job seekers had ever been called out for using AI.
Who is reading your application
The people screening applications are mostly recruiters who rarely present themselves as AI experts. In Refolk's index, about 122,600 people in the US hold Recruiter, Technical Recruiter, or Talent Acquisition Partner titles. Only 511 of them, about 0.4%, mention "AI" in their headline.
| Metric | Figure | Source |
|---|---|---|
| US recruiters / TA partners | 122,632 | Refolk's index |
| UK recruiters / TA partners | 9,364 | Refolk's index |
| US-to-UK ratio | 13.1x | Derived |
| US recruiters with "AI" in headline | 511 (0.42%) | Refolk's index |
| Hiring managers who say they can spot AI | 80% | Resume Genius 2026 |
| Hiring managers who'd reject an AI-generated application | 19.6% | TopResume 2025 |
The headline match is a free-text keyword search, so 511 is a rough lower bound. It still tells you something. The people judging whether your application "sounds like AI" are reacting to tone, not presenting themselves as AI specialists. That's good news if your resume says something specific, and bad news if it doesn't.
One caveat: TopResume surveyed hiring managers, not recruiters. Don't take 19.6% and apply it to the 122,600 US recruiters to get a head count of AI skeptics. Different population.
The real risk is sameness, not detection
The real problem with AI isn't that it gets you caught. It's that it makes every application look tailored, so tailoring stops telling the reader anything. A Yale study of 5 million cover letters across more than 100,000 Freelancer.com jobs shows how much signal disappeared.
Yale's Freelancer.com study found this even though only a minority of applicants used the AI tool.
The researchers, Jingyi Cui, Gabriel Dias, and Justin Ye, found that the correlation between how tailored a letter was and whether it got a callback fell by 51%. Only a minority of applications used the AI tool. That minority was enough to make tailoring worth half as much for everyone.
This is a signaling problem. A signal is any effort that's costly enough to show real intent. Tailoring a letter used to take real effort, which proved you cared about this specific job. When the cost drops to one click, it stops proving anything, and readers look for evidence that's still hard to fake. Employers on Freelancer.com shifted weight toward a freelancer's past work on the platform.
Inbound looks the same now. Pin's own survey says 92% of hiring teams see AI-written resumes in their inbound, and 43% say most of what arrives is machine-generated. Eva Chan of Resume Genius put it plainly: "As AI has raised the overall quality of resumes, it's also made them more uniform."
Recruiters can't catch AI. They can catch a bullet that could belong to anyone.
Why polished AI cover letters didn't win more interviews
AI improves the parts of a letter every candidate already has, and it barely changes the part recruiters care most about: why you, and why this job. In the Tilburg study, AI-assisted letters scored higher on average, and that didn't lead to more interview invitations.
The researchers found AI mostly improved the standard elements, while the motivation section stayed largely the same. Polish went up across the board and the field evened out. Till Wicker of Tilburg said candidates become more evenly matched, which makes it harder for recruiters to identify the best ones. Candidates who already wrote strong letters gained little, and they may even lose out if recruiters grow suspicious of AI use.
That's what applying into silence looks like now. Every letter you send reads cleanly. None of them gets a reply, because the applicants around you ran the same tool on the same posting and handed the recruiter the same well-written paragraph about being "excited to contribute."
Tilburg's broader finding was that AI worsens matching. When every letter sounds qualified, recruiters lose the ability to sort good fits from weak ones. The fix on your side is to apply where you actually fit and make the fit obvious.
How much AI help actually correlated with offers
Offers peaked with moderate AI help. In Jeremy Schifeling's Career Pulse survey of 240 job seekers, offer rates hit 13% for people who let AI write about half their materials and fell to 5% when AI wrote most of it.
| How much AI wrote | Got an offer |
|---|---|
| None | 7% |
| A little | 11% |
| About half | 13% |
| Most of it | 5% |
Two cautions: it's self-reported, and 240 people is a small sample. Even so, the shape is worth noticing. Skipping AI completely didn't help, and handing it everything hurt the most.
Schifeling's explanation: "When AI writes your story, you don't know your story." The cost shows up in the interview, not the screen. If a bullet you didn't really write gets you a call, the follow-up question exposes you.
The Yale paper points the same way. Time spent editing AI-generated cover letter drafts was positively correlated with hiring success. The draft is the cheap part. The editing is the part that still works as a signal.
More data is coming. Schifeling is presenting survey data on what recruiters really think about AI in hiring live with Jobscan on November 19.
What actually gives an application away as generic
What gives it away isn't an AI fingerprint. It's that nothing in the application could only have come from you. These are the patterns that make a skeptical reader assume AI, whether or not AI wrote it:
- Unnatural tone. Formal, even, a little grand. This was the top giveaway in the Resume Genius survey, named by 51%.
- Bullets with no nouns you own. "Drove cross-functional alignment to improve outcomes" fits every operator alive.
- The posting played back to itself. The job description's phrases repeated in order, with nothing added.
- Every bullet built the same way. Verb, scope, vague result, eight times in a row.
- A motivation paragraph that only names the company. It mentions the employer and says nothing about why your work leads there.
- Results without a baseline. "Improved performance significantly" doesn't let anyone check the claim.
How to write bullets that survive the screener and the skeptic
Write the kind of evidence the Yale employers moved toward: things a reference call or a portfolio could confirm.
- Name the system, client, or product. The internal tool, the account, the launch. A real proper noun can't be generated from the posting.
- Give a before and an after. "Cut monthly close from [X] days to [Y]" beats "streamlined close."
- Say what you decided, not just what happened. The tradeoff you made is the part a tool can't invent.
- Edit until you could defend every line out loud. If you can't explain a bullet in an interview, cut it.
- Tailor by choosing, not rephrasing. Pick the three experiences that match this posting. Don't rewrite one generic experience three ways.
This is the work Refolk takes off you without flattening you. I build the resume from your own history, so the bullets carry your systems, your numbers, and your clients. Then I tailor that same material to each posting and draft a cover letter whose "why here" points at something you actually did. You still do the editing pass, and you should.
FAQ
Is it OK to use AI to write my resume?
Yes, if you treat AI as a drafting tool and not as the author. In Schifeling's Career Pulse survey, the best-performing group used AI for about half the work (13% got an offer), the no-AI group did worse than light users, and the group that let AI write most of it did worst at 5%. The Yale Freelancer.com data points the same way: time spent editing AI drafts was positively correlated with getting hired.
How much should I edit an AI-drafted resume?
Edit until every bullet has something only you could have written, like a named system, client, or product, plus a before-and-after result, and until you could defend each line in an interview. The unnatural tone that 51% of hiring managers named as the top giveaway is exactly what a real editing pass removes.
Should I put a note on my resume saying I used AI?
There's no need for a disclaimer on the document. Resume Genius found 79% of hiring managers think candidates should disclose AI help, so in practice that means don't lie if someone asks. Answer plainly that you drafted with AI and edited every line against your own work history. If your bullets are specific and checkable, that answer won't cost you anything.