Quick answer
AI interview fraud detection is software and process that checks three things during remote hiring: that the candidate is a real person, that they are the person on the resume, and that they are answering on their own. It catches proxy interviewers, face swap deepfakes, stolen identities and hidden AI answer tools by matching ID to face, analyzing video and voice, and watching for screen overlays or location spoofing. Gartner expects 1 in 4 candidate profiles to be fake by 2028, so the safest setup pairs a detection tool with simple human checks and a human making every final call.
Key takeaways
- 91% of US recruiters in Greenhouse’s 2025 survey have spotted candidate deception, and 18% of hiring managers have seen a deepfake candidate.
- Fraud clusters in remote, well paid tech roles, where North Korean IT worker schemes have landed jobs at more than 100 US companies.
- Detection tools look at identity, video and voice, the candidate’s device and network, and how answers are produced.
- Cheap human checks still work: camera on, a hand wave, unscripted follow ups and an ID match before the offer.
- Biometric checks carry legal duties, including written consent under Illinois BIPA.
- Treat every alert as a reason to verify, never as an automatic rejection.
What is AI interview fraud detection?
AI interview fraud detection is the set of tools and checks employers use to confirm that a remote candidate is real, honest about who they are and doing the interview without hidden help. It sits on top of the video call or AI interview you already run. Most tools raise alerts during or after the interview; a recruiter then decides whether to verify further, continue or stop.
The term covers two different problems. Identity fraud means the person is not who they claim to be, from a stand in interviewer to a stolen identity. Assistance fraud means the right person is on the call but an AI tool is quietly feeding them answers. Good detection handles both, because they often show up together.
What types of interview fraud are recruiters seeing?
| Fraud type | What it looks like | How it gets caught |
|---|---|---|
| Proxy interview | A paid expert takes the interview; the real candidate shows up for the job | Face and voice matching across rounds, ID check at offer |
| Lip sync | The candidate mouths words while someone off camera speaks | Audio and lip movement analysis |
| Deepfake or face swap | A filter puts another face over the real person | Liveness checks, edge distortion, the hand wave test |
| Hidden AI answers | An overlay app listens and shows answers on screen | Device checks, answer pattern analysis, unscripted follow ups |
| Synthetic or stolen identity | A real person’s details, or a made up mix, on a fake profile | Document and liveness verification, duplicate contact details |
| Location spoofing | A VPN or remote laptop hides that the worker is overseas | Network signals, remote desktop detection, in person onboarding |
How common is candidate fraud in 2026?
Common enough that most recruiters have seen it. In Gartner’s survey of 3,000 job candidates, 6% admitted to interview fraud, either posing as someone else or having someone pose as them, and Gartner predicts 1 in 4 candidate profiles worldwide will be fake by 2028, as HR Dive reported. The numbers below come from employer and candidate surveys plus one vendor’s own applicant data.
| Finding | Source |
|---|---|
| 1 in 4 candidate profiles expected to be fake by 2028 | Gartner via HR Dive |
| 6% of candidates admit to interview fraud | Gartner candidate survey, 2025 |
| 91% of US recruiters have spotted candidate deception | Greenhouse 2025 report |
| 32% of hiring managers saw candidates reading AI scripts; 18% saw deepfakes | Greenhouse, same report |
| 36% of job seekers have used AI to alter their appearance, voice or background on video | Greenhouse, same report |
| 54% of recruiting leaders say AI made candidate fraud a much bigger problem | Gartner via Greenhouse |
| 16.8% of applicants to Pindrop were fake; 1 in 343 came from North Korea | Pindrop CEO in Fortune |
Keep the context in mind. The 36% figure includes candidates who used a background blur or a lighting filter to look more professional, which is not fraud. Pindrop sells fraud detection and was describing its own applicant pool. The direction is still clear: 74% of US hiring managers told Greenhouse they are more worried about fake credentials and deepfakes than a year earlier.
Why are North Korean IT workers part of this story?
Because they are the most organized version of it. In January 2025 the FBI warned that North Korean IT workers use stolen US identities, stand ins and face swapping technology in video interviews to get remote jobs, then steal data or extort employers, according to Ogletree’s summary. In June 2025 the Justice Department said these schemes had obtained jobs at more than 100 US companies, using US based “laptop farms” so overseas workers looked local.
How does AI interview fraud detection software work?
Most tools combine several signal types, because any single one is easy to fool. They score the session and flag specific moments for a recruiter to review.
- Identity: matches a government ID to the live face, with a liveness check so a photo or video loop fails.
- Video and voice: looks for face swap artifacts, lip movements that do not match the audio, and a voice that changes between rounds.
- Device: detects overlay apps, second screens, virtual machines and remote desktop software.
- Network: flags VPNs and locations that do not match the resume.
- Answers: spots long pauses before polished replies and language patterns typical of AI generated text.
Do fraud detection tools record the interview?
It depends on the tool, and the setup model matters as much as accuracy. Some join the call as a bot, some run inside Zoom or Teams with no bot at all, and some ask the candidate to install a small desktop app that checks their computer without recording audio or video. Ask each vendor what is captured, how long it is kept and whether any biometric data is stored.
Which AI interview fraud detection tools should you look at?
Pick by the fraud you actually see. Live interview overlays catch deepfakes, proxies and AI helpers on video calls; identity tools verify the person before an offer; AI interview platforms build fraud flags into their own screens. Prices come from each vendor’s own site, and most are quote based.
| Tool | What it catches | How it runs | Best for | Price from |
|---|---|---|---|---|
| Proxy, lip sync and deepfake signals | On your live video interviews | Teams that want a published entry price | $100 per month for 10 interviews | |
 Interview Guard | AI copilots like Cluely, overlays, VMs, VPNs, deepfakes | Candidate desktop app, no recording | Employers that will not record interviews | Quote based, 7 day trial |
| Deepfake, proxy, synthetic voice and ID mismatch | Agent joins Zoom, Meet or Teams | Mid market teams on Greenhouse, Lever or Ashby | Quote based | |
 VALIDIA | Face, voice and behavior mismatches, VPN use | Inside Zoom, Teams, Meet and Webex with no bot | Security minded remote hiring teams | Quote based |
 Polygraf AI | AI coached answers, deepfakes, synthetic voice | Live transcription and scoring, on premises option | Sensitive or government roles | Quote based |
| Fake or stolen IDs with liveness checks | Hosted ID verification flow | ID checks before an offer | $0.80 per check, $49 monthly minimum | |
 HireFoundry | Identity checks plus live interview guidance | Inside 30+ ATSs | Small teams wanting fraud checks in the ATS | $99 per month billed annually |
| Proctoring and integrity checks in video interviews | Its own interview and assessment platform | Enterprises running proctored assessments | Quote based | |
 HeyMilo | Cheating signals inside AI screens | Its own voice, video and text AI interviews | High volume teams using AI first rounds | Quote based |
Browse every vendor in our candidate fraud tools category. If AI answer tools are your main worry, our guide to real time interview assistants explains how Cluely style apps work and what they cost candidates.
How can recruiters spot a fake candidate without software?
Simple checks catch a lot. Recruiters on Reddit describe the same warning signs again and again: cameras that are “broken” every round, a voice that sounds different from the phone screen, long pauses before perfect answers, and sometimes a different person showing up on day one. These steps cost nothing.
- Require cameras on. Treat repeated excuses as a reason to reschedule, not to proceed.
- Ask for a hand wave. Waving a hand in front of the face can break face swap filters; one recruiter’s deepfake candidate left the call within seconds of the request, as HR Brew reported.
- Go off script. Ask follow ups on their own resume: what broke, who they worked with, what they would do differently.
- Ask local questions. The FBI suggests detailed questions about the candidate’s stated location and background.
- Check for duplicates. The same email, phone number or resume text across different applicants is a known red flag.
- Verify ID before the offer. Match a government ID to the face you interviewed, and call references through numbers you find yourself.
Are companies going back to in person interviews?
Some are, at least for a final round. Google, Cisco and McKinsey have brought back in person interviews to counter AI cheating and impersonation, according to Computerworld’s 2025 report. Greenhouse found 39% of US hiring managers are running more in person interviews to confirm authenticity, and Gartner found 62% of candidates are more likely to apply when a role includes an in person interview.
You do not need a fully in person process. A practical middle ground is remote screening with fraud detection, then one live step before the offer: an in person meeting, a verified ID video call, or in person onboarding with the laptop picked up locally.
What are the legal risks of fraud detection tools?
The biggest one is biometric data. Face matching and voice analysis can count as biometric collection, and Illinois’ BIPA requires written notice, written consent and a public retention policy, with damages of $1,000 per violation or $5,000 if reckless, as Paul Hastings explains. Get consent at the start of the interview and know where the vendor stores the data.
The second risk is false positives. Eye movement, pauses or a poor connection can trip alerts for honest candidates, including people with disabilities who may need an accommodation. Never auto reject on a fraud score. If the tool influences hiring decisions in New York City or Illinois, AI hiring laws may also apply; our AI interview statistics page covers those rules.
How do you build fraud checks into your hiring process?
| Stage | Check |
|---|---|
| Application | Flag duplicate emails, phones and near identical resumes |
| Phone or AI screen | Record consent; save a voice sample to compare later |
| Video interview | Cameras on, fraud detection running, unscripted follow ups |
| Final round | Same interviewer confirms it is the same person; in person if possible |
| Offer | ID and liveness check; references called on numbers you sourced |
| Onboarding | Ship equipment only to a verified address; watch for remote desktop tools |
Frequently asked questions about AI interview fraud detection
Can AI detect a deepfake in a job interview?
Often, yes. Detection tools look for face swap artifacts, lip movements that do not match the audio and failed liveness checks. No tool is perfect, so pair it with a hand wave test and an ID match before the offer.
How do I know if a candidate is using AI during an interview?
Watch for long pauses before polished answers, eyes tracking text off screen and answers that fall apart under follow up questions. Tools like Interview Guard and Polygraf AI flag overlay apps and AI coached answers directly.
What is a proxy interview?
A proxy interview is when someone else takes the interview for the candidate, usually a paid expert. The real candidate then starts the job. Comparing face and voice across rounds and checking ID at the offer stage catches most cases.
Is using AI in an interview always fraud?
No. Many employers allow AI for preparation, and some allow it in work samples. It becomes fraud when the candidate hides it in an interview that bans it or when someone else is answering. Tell candidates your rules in writing.
Which jobs see the most interview fraud?
Remote, well paid technical roles, especially software engineering and IT. They attract both professional proxy interviewers and organized schemes like the North Korean IT worker operations.
Do I need consent to run fraud detection on candidates?
Often, yes, especially for face or voice analysis. Illinois BIPA requires written consent for biometric data, so disclose the check before the interview and keep a record of consent.
Free newsletter
Get the Work Insiders weekly
One email a week: the recruiting and HR tech news that mattered, the tools we just added to the directory, and one practical guide. No spam, and you can unsubscribe in one click.