Guide
Can employers detect AI-generated resumes and cover letters?
Recruiters now see resumes and cover letters drafted with AI tools every day, and candidates want to know whether employers can tell. This guide explains what applicant tracking systems actually do, which signals recruiters notice, where AI detectors fit, and how to keep screening fair.
How common are AI-written applications?
Common, and mostly open. A Gartner survey of 3,290 job candidates in late 2024 found that 39% used AI during the application process; of those, 54% used it to generate text for a resume or CV and 50% to generate text for a cover letter (Gartner, 2025). Using AI to polish an application is not in itself dishonest; the risk for employers is fabricated experience, credentials or identity, which a writing detector cannot see.
Do applicant tracking systems detect AI resumes?
Not as a standard feature. Applicant tracking systems (ATS) are built to parse resumes into fields and rank candidates against job requirements. SHRM notes that traditional applicant tracking systems "are not designed to detect deepfakes or AI-generated resumes" and "primarily analyse skills, experience, and keywords" (SHRM, 2026). Jobscan, a resume-optimization company, states that "not one major applicant tracking system natively detects AI-generated resumes" (Jobscan, 2026).
Some employers add separate AI-detection or fraud tools to their hiring stack. Those tools carry the same limits as any detector: independent testing of 14 detection tools found them "neither accurate nor reliable" (Weber-Wulff et al., 2023).
What recruiters actually notice
This screening rubric separates signals that say something about a candidate from those that only say something about writing style. The weight column is AFIP's judgment, not a measured rate.
| What you see | Signal or non-signal | Why | Weight |
|---|---|---|---|
| Achievements with no names, numbers or context | Signal | Real experience usually leaves specifics a candidate can expand on | Moderate |
| Claims that conflict with dates, titles or references | Signal | Checkable facts; matters whoever wrote the text | Strong |
| Interview answers that cannot go beyond the resume wording | Signal | Tests knowledge directly, the best authorship evidence available | Strong |
| Polished, error-free writing | Non-signal | Many people use editors, templates and grammar tools | None |
| Phrasing similar to other applications | Weak signal | Templates and career-center advice also produce shared phrasing | Weak |
| A high AI detector score | Weak signal | Detectors misclassify human writing, including non-native English writing | Lead only |
Using an AI detector on applications
If you use a detector, apply it consistently and treat it as a reason to ask questions in the interview, never as a filter. Widely used detectors "consistently misclassify non-native English writing samples as AI-generated" (Liang et al., 2023), which turns a detector filter into a source of unfair rejections. A second detector that works differently, such as the free Walter Writes AI detector, can show whether a single score is an outlier; AFIP's guide on whether AI detectors can be wrong explains how to weigh the result.
Automated tools that score or rank candidates can also bring legal duties. New York City's Local Law 144 prohibits employers from using an automated employment decision tool "unless the tool has been subject to a bias audit within one year of the use of the tool," with audit information made public and notices given to candidates (NYC DCWP). Check with counsel before adding any automated screen; this guide is not legal advice.
An AI detector score should never decide whether a candidate gets an interview.
Fair, consistent screening
- State your position. Tell candidates in the job posting whether AI-assisted applications are acceptable and what must be the candidate's own work.
- Screen on substance. Score applications on verifiable experience and job requirements, not on writing style.
- Verify in the interview. Ask candidates to expand on specific achievements, tools and decisions from their resume.
- Check facts, not prose. Confirm dates, credentials and references; fabricated facts are the real risk.
- Treat everyone the same. If you run any check, run it on every application in the pool, not only on candidates who seem suspicious.
For related verification methods, see AFIP's guides on AI use in workplace reports and telling whether text was written by AI, or browse all AFIP verification guides.
Frequently asked questions
Can employers detect AI in resumes?
Not reliably with software. Major applicant tracking systems are not documented to flag AI writing, and standalone detectors make errors in both directions. Employers mostly notice AI-drafted resumes through vague achievements and through interview answers that cannot go beyond the resume, which is where verification works best.
Can companies detect AI cover letters?
Sometimes, by reading rather than by software. Generic letters that could fit any company stand out, and a detector score adds little on a one-page letter. A Gartner survey of 3,290 candidates found half of those who used AI applied it to cover letters, so recruiters see AI-assisted letters often.
Should candidates disclose AI use in a job application?
Follow what the employer asks. If a posting asks for disclosure or forbids AI-written material, comply. If it is silent, using AI to edit is widely treated like using any writing aid, but every claim must be true and you should be able to explain everything on your resume in an interview.
Is it fair to reject a candidate because a detector flagged their resume?
No. Detectors misclassify human writing, including non-native English writing more often than native writing, so a flag alone is weak evidence. A fair process uses any flag as a prompt for interview questions and decides on verified experience and skills.