⚡ Free forever — no login required

Detect deepfake candidates

Real-time face and voice overlays are now cheap enough that the person who passes your screen may not be the person who starts. These detection steps work today.

← All recruiting resources

How to detect deepfake candidates in interviews

The threat: Real-time face swap and voice cloning technology is now accessible enough that bad actors can impersonate candidates during video interviews. The person who passes your technical screen may not be the person who shows up on day one.
Detection: Ask them to turn their head. Most face-swap tools struggle with profile angles. Ask the candidate to look left, look right, or pick up their laptop and show you their workspace. A deepfake will glitch, lag, or refuse.
Detection: Watch for lip sync lag. If the candidate's lip movements are slightly out of sync with their audio (even by 100-200ms), that's a strong indicator of a real-time voice overlay.
Detection: Ask unexpected questions. Scripted deepfake candidates have prepared answers. Ask them something that requires real-time thinking about their specific past work: "Open your IDE and show me the last project you worked on." A real engineer can do this in seconds. A fake one can't.
Detection: Require camera on for the full interview. Some deepfake setups require the candidate to toggle their camera off and on to reset the overlay. If a candidate turns their camera off mid-interview and comes back looking slightly different, flag it.
Detection: Verify identity at onboarding. Require government-issued ID verification on day one that matches the person in the interview. Some companies now require a brief in-person meeting before the start date, even for remote roles.
Process fix: For high-sensitivity roles, add a short in-person or live-coding component to the interview loop. It doesn't have to be long. 30 minutes of screen-shared coding in a live IDE with real-time conversation is nearly impossible to fake.

How to verify candidate credentials

Cross-reference public signals: If a candidate claims to have built search ranking at Airbnb, check for public evidence: GitHub contributions, conference talks, blog posts, or mentions on the company's engineering blog. Senior engineers at top companies almost always have some public footprint.
Ask depth questions early: In the phone screen, ask specific technical questions about the work on their resume. "Walk me through the architecture of the system you built." Not "tell me about your experience." Depth questions expose fabricated experience immediately.
Verify employment directly: For senior roles, call the previous employer's HR department (not the reference the candidate provides) to confirm dates of employment and title. This catches candidates who inflate titles or extend employment dates.
Use your ATS data: If a candidate applied to the same company 6 months ago with a different resume, your ATS should flag the discrepancy. Cross-reference submissions from agencies against direct applications.

Where to Report Recruiting Fraud

Federal Trade Commission (FTC)

Report job scams and identity theft. The FTC tracks fraud patterns and takes enforcement action against repeat offenders.

reportfraud.ftc.gov →

FBI Internet Crime Complaint Center (IC3)

For internet-based fraud including phishing, identity theft, and employment scams that involve financial loss.

ic3.gov →

LinkedIn Fraud Reporting

Report fake job postings, fake recruiter profiles, and phishing messages directly to LinkedIn for investigation and removal.

LinkedIn Help Center →

Indeed Fraud Reporting

Flag suspicious job listings on Indeed. Their trust and safety team reviews reports and removes confirmed fraudulent postings.

indeed.com/trust-and-safety →

Common questions

How do I detect a deepfake in a live interview?

Ask for head turns and workspace shots (overlays glitch on profile angles), watch for lip-sync lag, and ask real-time questions like opening their IDE. Scripted fakes fail unscripted moments.

Why would someone use a deepfake candidate?

Proxy interviewing: a skilled impersonator passes the screen, and a different, cheaper worker takes the job, often with payroll routed through the fraud operator.

What is the strongest process fix?

A short live component that cannot be pre-recorded: 30 minutes of screen-shared coding with conversation, plus government-ID verification at onboarding matched to the interview recordings.

SourcingNav runs this playbook for you

Paste a job description and it builds the searches, scores every profile it finds, verifies emails, and drafts the outreach, with receipts for every score.

Paste a JD free →

More from the knowledge hub: Free Boolean Search Generator for Recruiters →X-Ray Search Guide →