Do In-Person Interviews Stop Hiring Fraud? Only One Kind

In-person interviews stop one kind of AI hiring fraud: impostors, including deepfaked and proxy candidates. They do far less against the more common kind, where a real candidate uses AI to fake skills they don't have. Six percent of candidates in a Gartner survey admitted to interview impersonation, while 64% of employers in a Criteria and Lighthouse Research & Advisory study said they had hired someone who misrepresented their skills. An in-person round is worth the cost for roles with access to sensitive systems. For most hiring, proctored assessments and structured interviews catch fraud earlier and for less.

Somewhere this fall, a recruiter will book a flight for a finalist whose resume, video screen and take-home exercise were all flawless. The candidate will show up, shake hands and turn out to be exactly who they said they were. The trip will have done its job. Whether it told the company anything about that person's ability to do the work is another matter, and usually the more expensive one.

Google, Cisco and McKinsey have each reinstated in-person rounds for some candidates, Computerworld reported, and Gartner found that 72.4% of recruiting leaders now conduct interviews in person to combat fraud. Gartner's definition of fraud in that figure is narrow: a candidate pretending to be someone else, or someone else sitting the interview in their place.

A conference room is good at catching that. Most hiring teams' bigger problem is something else.

Identity fraud vs. capability fraud

Identity fraud gets the headlines. Deepfaked video, cloned voices, proxies sitting in for the real applicant, and organized schemes like the North Korean IT worker operations the FBI keeps warning about. It's real, it's growing and it can be catastrophic when it puts someone inside a company's systems. As Talent Signal News has reported, a convincing fake candidate now takes about 70 minutes to build.

By the available numbers, though, it's the smaller share of the problem. In a Gartner survey of 3,000 job candidates, 6% admitted to impersonation, either posing as someone else or having someone pose as them.

Capability fraud is quieter and far more common. The candidate is who they say they are, and the skills are borrowed. In a separate Gartner survey, 39% of candidates said they used AI during the application process, and 29% of those used it to answer assessment questions. Criteria's 2026-2027 Hiring Benchmark Report, a survey of 1,027 U.S. hiring professionals, found that about 4 in 10 are struggling with candidates who use AI to misrepresent themselves, and more than a third can't reliably detect it. Earlier this year, Criteria and Lighthouse Research & Advisory surveyed 998 employers. Sixty-four percent had hired someone who misrepresented their skills, and 39% had done it more than once.

This is the fraud that ends up on payroll, and shaking someone's hand won't reveal it.

Joel Wolfe, president of the outsourcing firm HiredSupport, described what it looks like to Computerworld. Candidates were acing technical quizzes on an obscure JavaScript library that few developers know well. When hiring managers asked them to explain their answers, they couldn't. The giveaway was a follow-up question. That question works just as well over video, and it's useless in person if nobody asks it.

Gartner HR technology analyst Emi Chiba makes a fair point in the in-person round's favor: it's much easier to enforce an AI-free zone when the candidate is sitting across the table. True, but only if the conversation at that table is built to test something. An unstructured chat over coffee is one of the weakest predictors of job performance there is, whatever building it happens in.

What in-person interviews really cost

The line items are familiar. Airfare and hotels come back into the recruiting budget, scheduling drags out time to hire, and candidates in smaller markets or with caregiving duties quietly drop out. Talent Signal News covered those costs when Google, Cisco and McKinsey first made the shift.

Candidates aren't uniformly against it, to be fair. Gartner found 62% of candidates said they were more likely to apply if an employer requires in-person interviews, probably because it signals the job is real. The burden falls on the people who can't make the trip.

The bigger cost is timing. The in-person round almost always comes last. By the time a finalist walks into the building, every earlier call has been made on remote signals: who got screened in, who passed the assessment, who made the shortlist. If AI inflated those signals, the damage is done. The candidate with borrowed skills didn't just get through. They took the spot of someone who could do the job.

Researchers at Dartmouth's Tuck School of Business and Princeton modeled this. Once AI made polished, tailored applications nearly free, the most capable workers in their simulation were hired 19% less often and the least capable 14% more often, as Talent Signal News has reported. A final-round handshake can't undo that. It confirms the identity of whoever made it through a funnel that was already off.

How to catch AI hiring fraud earlier

Capability fraud gets in at the top of the funnel, so that's where to stop it. The tools for this have existed for years. They've just been sidelined while the resume did the screening.

A proctored, timed assessment covers both halves of the problem in one sitting. It confirms the person taking it is the person who applied, and it measures what they can do without help. Criteria, which sponsors Talent Signal News, builds its ability tests that way: strict time limits, large item banks and adaptive questions so no two candidates see the same test, plus aberrant response detection that flags coached or suspiciously perfect personality and risk answers. Its AI proctoring adds periodic photos, screen and audio sampling, and flags for second devices, tab switching or a different face partway through. Nobody has to fly anywhere, and it happens before the shortlist is set.

Employers already trust these tools more. In the Criteria and Lighthouse study, 98% rated assessments, structured interviews and work samples as more reliable than resumes, and only 2% said the resume is still their most trusted signal. Yet two-thirds still screen resumes first, a contradiction Criteria calls the resume trust gap.

Structured interviews pick up what's left. When every candidate gets the same job-related questions, scored against the same rubric, the follow-ups are built in and borrowed answers stand out. Josh Millet, Criteria's co-founder and CEO, has said structured interviewing is about twice as predictive as the unstructured kind, and that holds on video or in person.

Chris Daden, Criteria's chief technology and AI officer, has said there's no single fix. He advises assuming candidates are acting in good faith and telling them up front which uses of AI are allowed at each stage. Cisco has done exactly that. Scott McGuckin, its vice president of global talent acquisition, told Computerworld that AI is off-limits unless a candidate is explicitly invited to use it. Most companies have never written a rule down.

When an in-person interview is worth it

The in-person interview still has a job. It's a specialist tool, and too many companies are using it for everything.

It earns its cost for roles where identity is the risk. If a job comes with privileged access to systems, money or sensitive data, an impostor is far worse than a bad hire, and that's the situation the FBI's guidance on fraudulent remote workers was written for.

It's also useful as a check on everything that came before. Ask finalists to walk through how they approached the assessment or work sample they already completed. If someone scored well on a proctored test and can't explain their reasoning in the room, that gap is the clearest fraud signal a hiring team will get.

But it has to be structured, with the same questions, rubric and follow-ups for every finalist. Otherwise the company has paid for a pricier version of the least predictive interview there is. The same discipline works on a live video call, which is how Criteria built its real-time video interviewing: live conversations turned into structured, scored evaluations.

For everything else, spend the money at the top of the funnel. Verify identity and measure ability at the first real checkpoint, before AI-polished applications push capable people off the shortlist. By the final round, the interview should be confirming a decision the evidence already supports.

FAQ

Do in-person interviews prevent AI hiring fraud?

Partly. In-person interviews are effective against identity fraud, such as deepfaked video or a proxy taking the interview for someone else. They do much less against capability fraud, where a real candidate uses AI to misrepresent their skills earlier in the process. Capability fraud is the more common problem: 64% of employers in a 2026 Criteria and Lighthouse Research & Advisory study said they had hired someone who misrepresented their skills.

Why are companies going back to in-person interviews?

Mostly to stop candidate fraud. Google, Cisco and McKinsey have reinstated in-person rounds for some candidates, according to Computerworld, and Gartner found that 72.4% of recruiting leaders now conduct in-person interviews specifically to combat fraud. Deepfake video, voice cloning and real-time AI answer tools have made remote interviews easier to game.

What is capability fraud in hiring?

Capability fraud is when a real candidate uses AI or outside help to appear more skilled than they are, for example by generating resume content, answering assessment questions or reading AI-written answers during an interview. It differs from identity fraud, where the candidate is not who they claim to be. In Criteria's 2026-2027 Hiring Benchmark Report, about 4 in 10 hiring professionals said they are struggling with candidates who use AI to misrepresent themselves.

How common are fake or deepfake job candidates?

In a 2025 Gartner survey of 3,000 job candidates, 6% admitted to interview fraud, either posing as someone else or having someone pose as them. Gartner predicts that by 2028, one in four candidate profiles worldwide will be fake.

Are in-person interviews worth the cost?

For roles with access to sensitive systems, money or data, usually yes, because an impostor costs far more than a bad hire. For most other roles, proctored assessments and structured interviews catch more fraud, earlier and without travel costs. Any in-person round should be structured, with the same questions and scoring rubric for every finalist.

How can employers detect AI cheating in interviews?

Use structured interviews with consistent follow-up questions, and ask candidates to explain the reasoning behind answers they gave earlier in the process. Pair interviews with timed, proctored assessments that verify identity, and tell candidates in writing when AI use is and isn't allowed. A large gap between a candidate's assessment results and how they reason live is one of the strongest warning signs.