How to Screen AI-Generated Resumes Without Rejecting Good Candidates

AI-generated resumes have made every application look polished, and that has broken the old shortcut of judging candidates by how well they write. Recruiters now face more applications that read the same, a small share of outright fraud, and a real risk of rejecting honest people for using a writing tool.

This guide explains what the data says, why AI-text detectors are the wrong fix, and a five-check method that screens for evidence instead of prose.

Table of Contents

• What the data says
• Why AI detectors do not work
• The five-check method
• Signals that deserve a closer look
• Where tools help
• FAQ

What the data says

AI use in applications is widespread, but outright fraud is much rarer. Separating the two is the first step.

According to a Gartner survey:

• 39% of 3,290 candidates used AI during the application process. Of those, 54% used it for resume text and 50% for cover letter text.
• 6% of 3,000 candidates admitted to interview fraud, such as impersonating someone or having someone impersonate them.
• Gartner predicts that by 2028, one in four candidate profiles worldwide will be fake.

An SHRM article on fraudulent candidates adds that applicant tracking systems mainly screen for skills, experience and keywords, so they are not built to detect deepfakes or fabricated profiles. High volume also pushes recruiters to accept information without checking it.

So there are two different problems. Most applicants simply used a tool to write better. A small minority are misrepresenting who they are or what they have done. Your process should handle both without treating everyone as a suspect.

Why AI detectors do not work

It is tempting to run every resume through an AI-text detector. Resist that. Detectors guess from writing style, and good human writing and edited AI writing look alike. OpenAI withdrew its own AI text classifier because of its low accuracy.

There is also a fairness problem. Non-native English speakers and candidates who use editing tools for accessibility can be flagged unfairly. And if a tool scores or ranks candidates, it may fall under hiring-automation rules. See our guide to evaluating AI candidate screening software for the compliance questions to ask.

The five-check method

Instead of asking "was this written by AI?", ask "can this person show the work?" These five checks do that.

1. Look for specifics you can verify

Generic bullets such as "improved performance and collaborated across teams" say nothing. Look for named systems, measurable outcomes, links to repositories or products, and team sizes. For technical roles, public code and projects are strong evidence. Our guide to finding candidates beyond LinkedIn shows where those signals live.

2. Check consistency across sources

Compare the resume with the public profile, the application form and what the candidate says on a call. Mismatched titles, dates or employers are a better signal than writing style. Candidate enrichment can fill in and cross-check profile data automatically.

3. Hold a short live conversation early

A 10 to 15 minute conversation about one project, such as "what was the hardest decision, and what would you change?", is hard to fake and easy to compare. A structured AI interview can run the same questions with every applicant and produce a report, which keeps the process consistent when volume is high.

4. Use a small work sample for finalists

Give shortlisted candidates a short, role-relevant task and ask them to explain their approach live. The explanation reveals understanding that polished text cannot.

5. Verify identity before the offer

For remote hires, confirm identity and work history with reliable checks. SHRM recommends verifying employment and education directly, noting inconsistencies between resume and interview answers, and using in-person interviews where practical. Gartner's data supports this, since 62% of candidates say they are more likely to apply to a job that requires in-person interviews.

Signals that deserve a closer look

Signal	Why it matters	What to do
Vague achievements with no named systems	May be generic or fabricated	Ask for one detailed example on a call
Employment dates that overlap or do not match public profiles	Inconsistency	Verify with the employer
Candidate cannot explain their own project	Work may not be theirs	Run a live explanation or code walk-through
Camera or audio oddities, refusal to do any live step	Possible impersonation	Add an identity check and a second interviewer
Hundreds of near-identical applications	Mass-apply automation	Add a short screening question that needs a specific answer

A signal is a reason to look, not a reason to reject. Always give the candidate a chance to explain.

Where tools help

You cannot read 500 applications carefully, so use tools for the repetitive part and keep people in charge of decisions. For the high-volume playbook, read how to screen high-volume job applicants.

Good tooling does three things: it compares candidates against the same criteria, it pulls evidence from outside the resume, and it records why a candidate advanced. Whatever you use, keep a human reviewing rejections and keep records of how candidates were evaluated.

FAQ

Should I reject candidates who use AI to write their resume?

No. Many candidates use AI to edit and format, much like spell-check. Judge the evidence behind the claims, not the tool used to write them.

Can you detect an AI-generated resume?

Not reliably. Detectors are inaccurate and can flag honest candidates. It is more effective to verify the content through specifics, consistency checks and a live conversation.

How common is candidate fraud?

Gartner's survey found 6% of candidates admitted to interview fraud, and it predicts one in four candidate profiles could be fake by 2028. Rates vary by role and channel, and remote roles face more risk.

What is the best way to screen high volumes of applications?

Use consistent criteria, a short structured first conversation and automation for repetitive steps, then review finalists manually.

How do I spot a deepfake candidate in a video interview?

Look for visual or audio glitches, delayed responses and reluctance to do spontaneous tasks. Add identity verification, a second interviewer and, where practical, an in-person step before the offer.

Does this affect hiring-law compliance?

Tools that score or rank candidates may be regulated in some places, for example New York City's Local Law 144. Ask vendors about audits, keep human oversight and check with counsel. This is not legal advice.

Conclusion

You cannot out-detect AI-written resumes, but you can out-verify them. Screen for evidence, move live conversations earlier and reserve identity checks for finalists. That protects you from fraud without penalizing honest candidates.

Want every applicant interviewed the same way? Run an AI interview and compare structured reports side by side.