If hiring feels harder than it used to, here is a genuinely unsettling reason: a growing share of the people on the other end of your video interview are not who they say they are, and some are not even real. AI-powered interview fraud has surged through 2026, with detection tools now flagging 38.5% of candidates for suspicious behaviour, and the rate reaching 48% for technical roles. Gartner projects that by 2028, one in four job-candidate profiles worldwide could be fake. Recruiting is becoming a counter-fraud operation.
How the con works now
The toolkit has evolved past a browser tab of ChatGPT answers. Candidates now use real-time AI coaching feeding them responses through an earpiece, deepfake video that swaps the face on camera so the person you are interviewing is not the person who shows up, and smart glasses that discreetly display answers in the interviewer’s eyeline. “Proxy hiring,” where a skilled ringer sits the interview and a different, less-qualified person takes the job, has industrialised. The technical-role numbers are the scariest precisely because those jobs are remote, well-paid, and assessed through exactly the kind of screen-share tests AI is best at gaming.
This is the downstream cost of a technology that got very good at sounding competent very fast. When an AI can pass the interview, the interview stops measuring the candidate and starts measuring the candidate’s willingness to cheat, which is a different and much less useful thing to hire for.
What it means for anyone who hires
The practical fallout is that the remote-first, take-home-test, video-interview playbook that became standard over the last five years is now leaking badly, and companies are being forced to rethink it: live problem-solving that is hard to script, identity verification, in-person final rounds for sensitive roles, and assessment of judgement rather than recall. The bitter irony is that the same AI boom selling everyone productivity tools has also handed dishonest candidates a way to fake the very skills those tools were supposed to augment. If you hire, assume the process you trusted in 2023 is now gameable, and design for it. The goal has shifted from “find the best candidate” to “make sure the candidate is real, and actually the one doing the work.”
Related: the same problem is hitting classrooms.