For two decades the resume was the load-bearing data structure of hiring: a candidate produced one, software parsed it, and a recruiter ranked it. That structure is now buckling. New research finds that employers no longer trust the document they still build their screening process around, and the reason is that generative AI has flooded the top of the funnel with applications that look qualified whether or not the person behind them is. The hiring signal has degraded, and HR technology is scrambling to rebuild it somewhere other than the page.
The signal layer broke before the tools did
Research from Criteria Corp and Lighthouse Research and Advisory, based on a survey of 998 hiring leaders, captures the contradiction precisely. Only about one-third of employers say they are very confident that a resume reflects a candidate’s true skills, yet two-thirds still use the resume as the first screening step. The document has lost its credibility without losing its central position in the workflow.
The cause is not subtle. Some 92% of recruiting leaders in the study report that AI-generated resumes are now commonplace in applicant pools, and half describe them as very common. When a large language model can produce a polished, keyword-matched application in seconds, the resume stops measuring a candidate’s ability and starts measuring their access to a chatbot. That is a different variable, and it is not one any employer set out to hire for.
Advertisement
300 × 250
Why a polished document now hides the best people
The damage runs deeper than wasted recruiter time. The study points to an academic finding that when generative AI enters the application process, candidates in the top 20% of ability are hired 19% less often, while those in the bottom 20% are hired 14% more often. AI does not just add noise; it inverts the signal. It lifts weaker applicants to parity with stronger ones on paper, which means the screen that was supposed to surface talent now actively buries it.
Employers feel the consequence downstream. Nearly two-thirds say they have hired someone whose performance did not match the claims on their resume, and 39% say it has happened more than once. Teams that lean hardest on the resume as the primary decision driver are 35% more likely to report a bad hire. The document has become a liability dressed as a credential.
The shift: from parsing documents to measuring evidence
The vendor response reframes the resume as one weak input among several rather than the spine of the decision. Criteria positions itself as a talent signal platform, and the language matters: the product being sold is no longer resume parsing but evidence collection. Skills assessments, structured work samples, cognitive and behavioral measures, and validated simulations are being pushed earlier in the funnel, ahead of the human resume review rather than after it.
This is a structural change in where the screen happens. The traditional ATS workflow ranked documents and let assessments confirm a shortlist. The emerging workflow runs the assessment first and treats the resume as context, because the assessment is harder for a candidate to fake with a prompt. The point is not that assessments are immune to AI, but that a timed, structured measure of demonstrated ability is a sturdier signal than a self-authored narrative that a model can rewrite on demand.
The same AI sits on both sides of the desk
The pressure is symmetrical, which is what makes it hard. Candidates use AI to manufacture applications, and employers use AI to screen them, which produces a closed loop where one model writes to satisfy another model’s filter and neither output reflects a human. Separate research from Glean’s Work AI Institute, surveying 6,000 digital workers, found that for every hour employees spend producing useful AI output, they spend another hour making it usable, and more than a third of AI sessions fail outright. Applied to recruiting, that is a warning: automating the screen on top of an already-corrupted input does not restore the signal, it just industrializes the noise.
What this means for the HR leader
The instinct to bolt an AI resume-screener onto the existing funnel is exactly the wrong move, because it accelerates a process whose input is broken. The more durable response is to move the decisive measurement off the document. That means deciding which roles genuinely warrant a skills assessment or work sample before a human reads anything, and accepting that the resume becomes a context layer rather than a gate.
It also means treating candidate experience as a constraint. The same Criteria research found 68% of candidates would prefer that resumes be deprioritized in hiring, so a shift toward evidence-based screening is not a tax on applicants; many of them want it. The risk is assessment sprawl: a five-stage gauntlet that screens out strong people through fatigue rather than fit.
How to evaluate the shift
When a vendor pitches an AI hiring tool this year, the question is not whether it uses AI, because every tool now does. The question is which signal it actually measures. A screener that scores resumes more cleverly is optimizing a metric the research just showed is broken. A platform that puts a validated, job-relevant assessment in front of the resume is measuring something a language model cannot easily forge. Ask vendors to show the validity evidence behind their assessments, ask how they guard against AI-assisted cheating in remote testing, and insist on adverse-impact data before any tool touches a live req. The resume is not disappearing. But the organizations that come out of this ahead will be the ones that stop asking the document to carry weight it can no longer bear.
Source: Criteria Corp via PR Newswire
