Resume Fraud Is Real. The Interview Gap Is What Lets It Through.

A resume comes in. It's clean. Metrics sit in the right places, the job titles escalate in a logical line, the language mirrors the posting almost too closely. Nothing on it is technically wrong. It just feels manufactured — and there's no longer a reliable way to tell if that instinct is right, because most resumes now feel that way.

That's not a hunch. It's the new baseline. A recent Cadient analysis of 3,000 resumes found AI-generated content in roughly three out of four applications for higher-salaried roles, and in half of applications for hourly ones. Nine in ten of the flagged resumes contained outright hallucinations: overlapping employment dates, promotions that don't fit a real timeline, career arcs that read more like fiction than history. Government and real estate hiring came out worst, at a 44% flag rate. This isn't an edge case anymore. It's the median application, and it's a legitimate risk on its own terms — bad data at the top of the funnel means bad decisions downstream, regardless of what happens later in the process.

The pattern holds well beyond one report. Gartner projects that by 2028, roughly one in four candidate profiles worldwide will be fabricated in some way. Robert Half found that two-thirds of hiring managers now say AI-generated applications are actively slowing hiring down, not speeding it up, because everything looks equally polished and nothing is easy to trust. Greenhouse's research puts close to a third of recruiters' week into filtering junk applications alone. The volume of talent didn't go up. The volume of noise that resembles talent did.

Two problems, not one

The instinctive response to "AI is generating convincing fake resumes" has been "build AI that detects convincing fake resumes" — extensions that cross-reference LinkedIn, semantic matching instead of keyword matching, identity verification at the top of the funnel. That work matters, and none of it should be deprioritized. A synthetic resume that gets past screening wastes recruiter hours, distorts pipeline metrics, and in the worst cases opens the door to identity fraud or state-sponsored infiltration schemes that have nothing to do with interview quality. Resume-layer verification is a real control, not a decoy.

The problem is treating it as the only control. Proxy interviewers, coached real-time answers, deepfaked video — none of that is built to beat a resume parser. It's built to survive whatever comes next, because for most organizations the interview is still the least structured, least defensible stage in the entire process. A resume that clears screening just earns a seat at the table. What happens at that table is supposed to be the real evaluation, and for most companies it still runs on instinct, not rigor. Catching fraud at the document is necessary. It was never sufficient — even before AI made fabrication this easy.

That's the actual gap: hiring teams spent the last few years reinforcing the front door and left the room behind it essentially unchanged. Both need to hold at once.

What actually predicts whether someone can do the job

This isn't a new idea, and that's exactly what makes it useful. Decades before generative AI existed, Frank Schmidt and John Hunter published the meta-analysis that's still the reference point in personnel psychology today: structured interviews — same core questions, same scoring rubric, applied consistently across every candidate — predict job performance at roughly double the validity of the free-flowing, conversational interviews most companies still default to. Later research from McDaniel and Whetzel found the same pattern holds across seniority levels and industries. Google's own internal hiring research landed in the same place.

The mechanism is almost mundane: structure removes the room for a rehearsed narrative to substitute for demonstrated experience. A candidate can prepare a compelling line about "leading cross-functional alignment." It's far harder to improvise specific, situational follow-ups about a project that doesn't actually exist, scored against a rubric that wasn't built on the fly. Unstructured interviews reward performance. Structured ones reward evidence.

Which makes the timing notable. The exact discipline that's sat underused in I/O psychology research for decades — because it takes more work to build and run consistently — is now a direct answer to a problem the research never anticipated. AI didn't just make resumes easier to fabricate. It raised the stakes on the interview being the stage where fabrication finally has to hold up under specific, real pressure.

Verification and depth, together

The strategic response to synthetic resumes isn't choosing between better screening and better interviewing — it's not treating either one as sufficient on its own. Most talent acquisition teams don't have the appetite to run identity verification and deepfake monitoring on every applicant, and most roles don't warrant that level of scrutiny. But every team can control what happens once a candidate is in the room: an interview rigorous enough that a fabricated background can't survive it, not because someone got caught lying, but because the questions were specific enough that only real experience could answer them well.

That's the problem InterviewIQ, from HR Success Centre, is built to solve. It isn't a resume screener or an AI detector racing an arms race it can't win — it's a structured interview guide generator: competency-mapped questions, anchored scoring rubrics, and follow-up prompts designed to move past the rehearsed version of an answer into the specific, situational detail only someone who actually did the work can produce. The same mechanism Schmidt and Hunter documented in 1998, applied to a hiring environment they never could have anticipated.

Resume verification catches what gets submitted. A strong interview catches what can't be faked in the room. Hiring teams need both — but only one of those two stages has been getting the investment.