For most of the history of professional hiring, the interview did double duty. It assessed capability, and — almost incidentally — it proved that the person describing the capability was the person who possessed it. Those two functions have now come apart, and the second one is failing quietly.
Gartner
Gartner's projection covers a spectrum rather than a single behaviour. At the mild end sits cosmetic enhancement: a language model turning an unremarkable career into compelling prose. At the severe end sits substitution — a different person sitting the interview, coached in real time or rendered through synthetic video. Pindrop's 2025 Voice Intelligence Report recorded a thirteen-fold year-on-year increase in deepfake attempts within hiring processes. In one survey of 3,000 candidates, six percent admitted to some form of interview fraud outright.
Why senior hiring feels safe, and isn't
The instinctive response from leadership hiring is that this is a volume-recruitment problem. Fraud at scale targets remote engineering roles and offshore contract work, where the hiring process is thin and the hire is never seen in a room. Executive search, the argument goes, involves too much scrutiny for a fabricated candidate to survive.
That is half right, and the wrong half is the dangerous one. Outright impersonation is genuinely rare in senior search. But the milder failure — a career narrative inflated past what the record supports — has become dramatically easier to produce and dramatically harder to detect by reading. A CV is now a generated artefact. So is a cover note, a LinkedIn summary, and increasingly the first-round answer to a competency question. The document no longer carries evidentiary weight, because producing a polished one costs nothing.
The written record has stopped being evidence. It is now a claim, and claims have to be checked against people rather than paper.
The defences that actually hold
Verification technology will improve, and it should. But the durable defence in leadership hiring was never technological, and it predates the problem. It is that senior markets are small, and reputations in them are held by people, not documents.
Three practices do most of the work:
- Back-channel referencing. Not the referees a candidate nominates — the people who worked alongside them, found independently. A nominated referee validates a story. A back-channel reference tests whether the story is the one their former colleagues would tell.
- Provenance over description. Rather than asking what someone achieved, establish who else was in the room when it happened, what the counterfactual was, and what they would have done differently. Fabricated narratives are thin behind the headline; real ones have texture and, usually, regret.
- Continuity of relationship. A candidate known to a search partner over several years, through two or three moves, carries a verification history no single process can reconstruct under time pressure.
The uncomfortable second-order effect
There is a failure mode on the other side of this that deserves naming. As detection anxiety rises, organisations reach for verification tooling that penalises the wrong people — candidates who write in a second language, who used assistive tools legitimately, or whose careers are non-linear and therefore read as improbable. Fraud detection that runs on stylistic suspicion will systematically disadvantage exactly the candidates a widened talent pool was meant to include.
The distinction that matters is not whether AI touched the document. It is whether the underlying claim is true. Those are different questions, and only the second one is worth building a process around.