London, November 2025
For decades, hiring ran on a chain of trust: a CV claimed a skill, an interview tested it, and a reference confirmed it. In 2025, generative AI snapped every link in that chain at once. Candidates can now generate a flawless application in seconds, rehearse with an AI coach, and — increasingly — pipe AI answers into the interview itself. The result is a quiet crisis of trust at the heart of graduate recruitment: employers can no longer be sure that the person on paper is the person who will turn up to work.
The data is stark. According to the Institute of Student Employers (ISE) — the UK's authoritative early-careers body — 48% of employers are now concerned that graduates are misrepresenting their abilities using AI. The share of employers who say they frequently encounter cheating in their selection process has doubled, from 7% to 15%. And most tellingly, 61% of the misconduct employers reported involved candidates using AI during interviews, without disclosing it. The conclusion, says Ewance — a challenge-based learning-and-recruiting platform where students solve real-world challenges tied to their field of study, earn verifiable certificates, and recruiters discover and verify proven talent — is that the signals hiring has always relied on are quietly losing their meaning.
Half of employers no longer trust the CV
The erosion of trust is not a fringe worry; it is approaching the recruiting mainstream. With nearly one in two employers doubting the authenticity of graduate self-presentation, the foundational documents of hiring — the résumé, the cover letter, the personal statement — are losing their value as filters precisely because they are the easiest things for AI to produce.
This matters most at the entry level, where candidates have the thinnest track record and the system leans hardest on self-reported potential. A graduate with no work history has always been, in effect, a promise. AI has made that promise trivially easy to embellish — and employers know it.
The CV was already a weak signal. AI has finished the job. When anyone can generate a perfect answer to any question, the answer stops being evidence of anything. Employers are realising they have been screening for the ability to use AI — not the ability to do the job.
— Ewance's founder
The interview was the last line of defence. AI crossed it.
If the CV was always imperfect, the interview was supposed to be the backstop — the live, human moment where claims met reality. That backstop is now failing too. The ISE's finding that 61% of reported misconduct involved undisclosed AI use during interviews is the single most alarming number of the year for recruiters, because it means the verification step itself has been compromised.
Once the interview is compromised, employers are flying blind. They are making multi-year bets on people based on a conversation that may have been ghost-written in real time. No serious organisation can build a workforce on that.
— Ewance's founder
Are recruiters rebuilding the process mid-flight?
Employers are not standing still. The same ISE research shows that 79% of employers are redesigning or reviewing their recruitment processes in response to candidate AI use, with a full third redesigning specifically because of AI. Looking ahead, 62% expect to use AI in their own recruitment within five years — fighting fire with fire.
But redesigning the process does not, on its own, solve the trust problem. Adding AI to screen out AI is an arms race, and history suggests the defenders rarely win it for long. The deeper question employers are now asking is not "how do we detect AI?" but "what can we measure that AI cannot fake?"
Why detection is a losing game
It is tempting to reach for a technological fix — AI-writing detectors, proctoring software, ever-tighter controls. The trouble is that detection treats the symptom. Detectors are gameable, generate false accusations that punish honest candidates, and are always one model release behind the tools they chase. Every hour a recruiter spends adjudicating whether an answer was human is an hour not spent assessing whether the candidate can actually do the work.
The problem compounds because the demand for a reliable signal has never been higher. Separately in 2025, LinkedIn — a global, interested-party source — reported that 89% of talent-acquisition professionals say measuring the quality of a hire is increasingly important, yet only 25% feel confident they can actually measure it. Employers are, by their own admission, trying to make better decisions with worse information.
So what can't AI fake?
There is one signal generative AI cannot counterfeit: the demonstrated act of solving a real problem. AI can write an essay about teamwork; it cannot have done the project. It can describe a marketing strategy; it cannot show the work of having built and defended one against a real challenge. The closer assessment moves to authentic, applied work — and the further it moves from self-report — the less room there is to fake it.
This is why the most durable answer to the AI trust crisis is not more policing, but better proof. Evidence of competency earned by actually completing real-world challenges, tied to a candidate’s field of study and independently verifiable, restores exactly the signal AI has eroded: a credential a recruiter can trust because it reflects work that was genuinely done.
The market is converging on a simple idea. Stop asking candidates to describe their skills, and start letting them demonstrate them on something real. Proof you can verify beats a claim you have to take on faith — and in an AI world, that is the only kind of evidence left standing.
— Ewance's founder
What this means heading into 2026
For employers, the lesson of 2025 is that selection has to shift from claims to evidence — from what a candidate says to what a candidate has verifiably done. For students and graduates, it is that the surest way to stand out in a sea of AI-polished applications is to carry proof that cannot be generated: a record of real problems solved.
Recommendations
Ewance's founder offered the following for employers and graduates navigating the trust crisis:
- Assess work, not words. Anchor selection in authentic tasks AI can't complete on a candidate's behalf.
- Prize the verifiable. A credential a recruiter can independently check is worth more than a claim that must be trusted.
- Stop fighting an arms race you'll lose. Detection chases the symptom; demonstrated proof of skill removes the incentive to fake.
- For graduates: build evidence, not just applications. In a market flooded with AI-written CVs, proof of real work is the differentiator.
AI did not break hiring. It exposed how much of hiring was running on trust we could no longer afford. The organisations that come out ahead will be the ones that stop asking to be told what a candidate can do, and start asking to be shown.
— Ewance's founder
Sources
All figures are drawn from the original sources below; Ewance conducted no proprietary survey for this release.
- Institute of Student Employers (ISE), Student Recruitment Survey 2025 / "Top 10 stats of 2025" — 48% of employers concerned graduates misrepresent abilities using AI; employers frequently encountering cheating doubled from 7% to 15%; 61% of reported misconduct involved candidates using AI during interviews without disclosure; 79% of employers redesigning or reviewing recruitment due to candidate AI use (one-third specifically because of AI); 62% expect to use AI in recruitment within five years. UK.
- LinkedIn, Future of Recruiting 2025 — 89% of talent-acquisition professionals say measuring quality of hire is increasingly important, while only 25% feel confident they can measure it. Global, vendor-sourced (interested-party); figures are attitudinal.
