AI didn't invent resume fraud. It just industrialized it.

People have always shaded the truth on a resume. What changed is the scale, the speed, and in some cases, whether the person in the interview is even real.

The numbers behind the discomfort

Roughly six percent of candidates admit to interview fraud outright, either posing as someone else or having someone else stand in for them. That's a small number on its own, but recruiters aren't seeing it that way. In recent industry surveys, the large majority of recruiters and hiring managers report having spotted or suspected candidate deception in the past year, and most say they're more worried about fake credentials now than they were twelve months ago.

Some of what's driving that worry is mundane: a cover letter that oversells a skill, a resume timeline stretched to close a gap, but some of it is stranger. Reported cases now include recruiters conducting entire interviews with a deepfaked candidate before anyone noticed. That's not the common case, but it's a real one, and it's the kind of story that makes every recruiter reread their last three interviews a little differently.

Self-report was always the soft spot

None of this is a new vulnerability. It's an old one, revamped, and running at a new speed. People describe themselves the way they'd like to be seen, and that pull gets stronger whenever the answer has consequences. A resume and a cover letter are entirely self-report, written by the candidate, about the candidate, with no one else in the room to check it against anything.

What generative tools added wasn't a new kind of dishonesty, but scale. A resume that used to take an evening to polish now takes a prompt. A cover letter tailored to one specific role used to be a rare effort; now it's the default, sent to fifty postings in an afternoon. The soft spot got bigger, and much easier to lean on.

What gets harder to fake

Amid all this, there's the useful distinction, and it's a narrow one: written application is built entirely from self-description: how someone chooses to present themselves, with as much time as they want to get the wording right. A live conversation is a different kind of input. Someone talking through an actual problem, in real time, isn't the same task as writing a polished paragraph about themselves, and it's considerably harder to script convincingly at scale.

We're not saying a conversation catches every dressed-up application, or that it functions as a lie detector. What we are saying is that the easiest thing to polish is the thing you had unlimited time and zero witnesses to write.

The takeaway

Looking better on paper than in practice has always paid off for candidates willing to do it. AI mostly changed the price: what used to take real effort now takes a prompt, and in the rarest cases, the candidate isn't even a real person. Weighting a process less on the material that was always easiest to embellish matters more now than it did five years ago, not because embellishment is new, but because it got this cheap.