Can AI detectors tell if a resume was written by AI?
Not reliably, and the best-known attempt was withdrawn. OpenAI's AI text classifier, launched January 31, 2023, caught 26% of AI-written text in OpenAI's own tests and wrongly flagged human writing 9% of the time. OpenAI took it down on July 20, 2023, "due to its low rate of accuracy."
Detectors also misfire on some human writers more than others. A 2023 study in the journal Patterns by Liang and colleagues ran 91 essays by non-native English speakers through seven detectors. On average, the detectors labeled 61.3% of those human-written essays as AI-generated, while they classified essays by US eighth graders accurately.
Resumes make the problem harder. They are short, and OpenAI said its classifier got more reliable as text got longer. They are also written in a clipped, formulaic style that people used long before AI. We found no published study that tests AI detectors on resumes specifically, so any claim that a recruiter's software can spot an AI resume has no public evidence behind it.
Do employers mind if AI helped write your resume?
Most large employers that have published a rule allow AI help with wording and ban AI that invents things. McKinsey's careers page, checked October 22, 2026, marks "Polishing your resume" as encouraged. It marks "Exaggerating or creating achievements that aren't yours" as not allowed. Accenture's page draws the same line.
Accenture's page on AI in its hiring, also checked October 22, 2026, says "AI can help with grammar, tone, and structure" and adds "Make sure your final version reflects your real story and unique experience." It forbids "Using AI to generate false or misleading experiences, skills or qualifications on your resume or cover letter." The Interviewing at McKinsey page says: "We want to get to know you, not your AI tool."
Amazon went further and published its own guide. In December 2024 its news site ran a guide to writing your resume with generative AI, using an AI tool Amazon built to compare a resume with a job posting. So the employer's test is whether every claim on the page is true. The rules at five large employers are quoted in full on the page about employer AI interview policies.
Does an AI-polished resume hurt your chances?
The best evidence says no, as long as the content is yours. A field experiment by Wiles, Munyikwa and Horton, published as an NBER working paper in 2023, gave nearly half a million jobseekers on an online labor market algorithmic writing help on their resumes. The jobseekers who got the help were hired 8% more often.
The authors also checked the obvious worry, that cleaner writing would hide weaker workers and disappoint employers later. They found no evidence that employers were less satisfied with the people they hired. The help changed how clearly people described their work, and the work itself stayed the same.
That study tested writing assistance on an online labor market, so it may not carry over exactly to a corporate job search. It still points the same way as the employer rules above. Better wording around true facts helps, and the risk lives in the facts.
What do recruiters actually check on a resume?
They check your claims, mostly by asking about them. Interviewers question the bullets on your resume, assessments test the skills you listed, and background checks can confirm your jobs and degrees. The table below lists each check, when it usually happens, and the source for it. None of them depend on spotting AI wording.
| Check | When it happens | What it tests | Source |
|---|---|---|---|
| AI text detector | Rarely, and with known errors | Whether text looks machine-written | OpenAI withdrew its classifier in July 2023 for low accuracy (OpenAI) |
| Questions about your bullets | Every interview round | Whether you can explain how you got each result | Google's vice president of recruiting said interviewers probe answers to test real understanding (CNBC, 2025) |
| Skills assessment | Before or between rounds | Whether you have the skills you listed | McKinsey and Accenture ban AI during assessments unless allowed (McKinsey, Accenture) |
| Background check | Usually near an offer | Employment history and education | US employers may ask about both, and need your written permission for a report from a screening company (EEOC and FTC) |
| Employer AI rules | At application or interview | Whether you invented or exaggerated achievements | McKinsey lists "Exaggerating or creating achievements that aren't yours" as not allowed (McKinsey) |
The background check row comes from a joint EEOC and FTC guide for job applicants. It says employers may ask about "your employment history, your education" and more, and that an employer "must ask for your written permission" before getting a background report from a company that compiles them. That is US guidance. Other countries have their own rules.
Why is an invented number the real risk?
Because it's the one thing on an AI-polished resume that someone will test. A number invites a question: how did you measure that, over what period, against what baseline? If an AI tool wrote the number for you, you have no answer, and the interviewer now doubts everything else on the page.
Here is how it happens, as an example. You give a tool a bullet that says "ran the onboarding redesign." It comes back as "led an onboarding redesign that improved activation by 40%." You never measured activation. The sentence reads well, and nobody would flag it as AI. In the interview, the first follow-up asks where the 40% came from.
That is the unverifiable-claim problem: a claim you can't support exposes you, however the sentence was written. Any tool that fills a gap with a plausible figure creates this risk. Why that happens, and how to stop your tool doing it, is covered in why AI resume tools invent numbers.
How do you use AI on a resume without the risk?
Lock the facts before the AI writes anything. List each result you can defend, with the number and where it came from, and let the tool reword only those. If a line needs a figure you don't have, leave the figure out. A shorter bullet you can explain beats a longer one you can't.
This is sometimes called fact-locking. The AI can change the verb, the order and the length. It can't add a number, a title, a team size or a tool you never used. Run every bullet through three questions before you send it:
- Did this happen the way the line says?
- Could you explain in one minute how the number was measured?
- Would a former manager confirm it if asked?
Any bullet that fails one question gets rewritten or cut. That takes about as long as reading the resume aloud once.
Prept's resume side is built on that rule: every claim on a tailored resume traces to a fact you confirmed, and numbers you can't stand behind are left off rather than invented. The resume optimizer shows how the confirmation step works, and the free tier reviews one resume and tailors one application with no card needed (try it free).
Should you tell a recruiter that AI helped with your resume?
You don't need to volunteer it for wording help, since the employers quoted above allow polishing. If someone asks, say yes plainly and say the facts are yours. What you need to be ready for is a question about any claim on the page, and that readiness matters more than any disclosure.
After each interview, note which resume lines the interviewer asked about and how well you answered. The free 10-minute debrief template has room for it. Lines that drew a weak answer are the ones to rewrite before the next application. The page on whether to tell an interviewer you used AI to prepare gives the one sentence to use, and where using AI in an interview becomes cheating draws the wider line.