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

Watch for lag between mouth and audio, flat lighting that never changes, and eyes that do not track naturally when you move on screen. Ask the candidate to turn their head sideways or hold a hand up next to their face. Most current deepfake tools break on profile angles and hand-to-face motion. Also ask a specific, off-script question about a project on their resume. Fakes stall or give vague answers because there is no real person with real memories behind the profile.

Who is liable if my firm places a fraudulent worker at a client?

You are. The client hired your firm to vet the person, so the failure lands on you first. Depending on your contract and the damage, you could face breach claims, indemnification demands, and a lost account. If that worker touches sensitive data or systems, you may also be pulled into a security incident. This is why verification is a business development issue, not just an HR checkbox. Clean placements are now a selling point.

Will tighter verification slow down my time-to-fill?

A little, if you bolt it on badly. Almost none, if you build it into the workflow. The trick is to run cheap checks early and expensive checks only on finalists. A quick liveness prompt and ID match at first screen takes under two minutes. Deep verification runs only on the one or two people you actually plan to submit. Done right, you lose minutes and save yourself from a placement that blows up in month two.

AI Resume Fraud: Rebuild Candidate Verification Now

AI-generated identities and deepfake interviews are now the top hiring challenge TA leaders name for 2026. If your firm places a fake worker into a client's systems, you own the liability. Here is the verification playbook to fix it this quarter.

Gartner projects that by 2028, one in four candidate profiles will be fake. Not padded. Not exaggerated. Fake. Built by AI, wrapped in a synthetic identity, and pushed through your pipeline by someone who never plans to show up as the person on the resume.

TA leaders now name this as the number one hiring challenge for 2026. And for staffing firms, it is worse than for corporate recruiters. When you submit a candidate and the client hires them, you are the vetting layer. If that worker turns out to be a fraud sitting inside the client's systems, you carry the liability and the reputational hit. The client remembers who sent the fake.

This is not a future problem. It is happening in your pipeline right now, and most firms are still verifying candidates the way they did in 2019.

What AI resume fraud actually looks like in 2026

Forget the old picture of a slightly exaggerated resume. The new fraud runs deeper and moves faster.

Synthetic identities. A real-looking person with a real-looking work history, a LinkedIn profile, and references who are also fake. None of it traces back to an actual human.
Deepfake interviews. A live video call where the face and voice are AI-generated in real time. The person answering questions is not the person who will do the job.
Proxy candidates. One skilled person interviews. A different, cheaper, less-skilled person shows up to work. Common in remote IT and offshore placements.
AI-written resumes tuned to your ATS. These are not fraud by themselves, but they hide weak candidates behind perfect keyword matches. Your screen passes them because the resume was built to pass your screen.

The tools to build all of this cost less than a monthly software subscription. That is why the volume is climbing so fast.

Why your current process misses it

Most staffing firms verify at the end. Background check after the offer. Reference calls after the client says yes. Skills assessment maybe, maybe not.

That order made sense when fraud was rare and expensive to pull off. It fails now for two reasons.

First, background checks confirm that an identity exists. They do not confirm that the person on your video call is that identity. A synthetic profile with a stolen or manufactured Social Security number can clear a standard check.

Second, everything happens too late. By the time you find a problem, you have already spent the effort, made the submission, and put your name behind the person. Pulling a candidate after submission costs you credibility with the client every single time.

The fix is not more checks. It is checks in the right order, run at the right point, on the right people.

The verification playbook: build it in three layers

Think of this as Build. You are building a system that catches fraud early, cheaply, and without wrecking your time-to-fill. Run it in three layers.

Layer 1: Cheap checks on everyone, up front

These run at first contact and take minutes.

Liveness check. On the first video screen, ask the candidate to turn their head to the side and hold a hand next to their face. Real people do this without thinking. Deepfake tools break on it.
ID match. Ask for a government ID on camera and compare the face live. Free tools do this well enough for a first pass.
One specific question. Pull one detail from the resume and ask about it in a way that requires a real memory. "Walk me through the hardest week on that project." Fakes give you fog.

Layer 1 costs you under three minutes per candidate and knocks out most low-effort fraud before it enters your pipeline.

Layer 2: Deeper checks on your finalists

Run these only on the one or two people you plan to submit.

Live skills demonstration. No take-home tests. Watch them solve a real problem in real time, on camera, screen shared. Proxy candidates fall apart here.
Direct reference contact. Call references through numbers you find yourself, not numbers the candidate hands you. Fake references are usually supplied by the fraudster.
Employment verification through the source. Confirm past employers exist and confirm the dates through the company, not through a document the candidate provides.

Layer 3: Client-facing proof

This is where verification becomes business development. Package what you did.

Tell the client, in plain terms, what your verification process catches.
Give them a short summary of the checks you ran on the person you submitted.
Make clean placements part of your pitch. In 2026, "we verify identity with live liveness and skills checks before we ever send you a name" is a reason to pick you over the firm that does not.

Your competitors are treating verification as a cost. Treat it as a differentiator and you win accounts on it.

The mistake to avoid

Do not buy one expensive fraud-detection tool and call it done. Tools help, but the fraud changes faster than any single tool keeps up. What holds up is the layered process, run by trained recruiters who know what a fake looks like.

Train your team on the three Layer 1 signals this month. That alone catches more fraud than any software you can buy this quarter.

One more thing for PE-backed portfolios. Standardize this across every brand in the portfolio. One fraudulent placement at one operating company can taint the whole platform's reputation with a shared client base.

Your move this week

Pick your five most recent submissions to a sensitive client, meaning anyone placed near data, money, or systems. Run Layer 1 checks on each one retroactively where you can, and score how confident you are that each person is real. If you cannot say yes for even one of them, you have your answer, and you have your starting point.

Then add the three Layer 1 steps to your first-screen script and use them on every candidate starting Monday. It costs you three minutes and protects the one thing you cannot rebuild fast: your name with your clients.

The AI Resume Fraud Playbook: Rebuild Candidate Verification Before It Becomes a Client Crisis