92% of UK Graduate Employers Report Interview Misconduct as AI Cheating Surges

Ninety-two percent of UK graduate employers suspect or confirm candidate misconduct during hiring this year — a record high — and the most common offence is undeclared use of AI in interviews and online assessments.

The Institute of Student Employers (ISE), the body representing 290 graduate employers including Marks & Spencer, Rolls-Royce and Accenture, reported the figure in its annual survey. The number has climbed steadily: 85% of employers reported suspected or confirmed misconduct last year, and 78% the year before.

Two-thirds of respondents — 66% — reported at least one incident of a graduate misusing AI during an interview. That makes unauthorised AI use in interviews and online assessments the single most common form of misconduct in the graduate hiring process. The ISE did not name which of its member employers had encountered AI-assisted interviews.

The finding lands at a moment when AI's effect on the labour market is the subject of intense debate. Employers have widely been expected to cut entry-level roles as automation reduces the need for human workers, making the integrity of the hiring funnel — the mechanism that sorts graduates into those jobs — a matter of both market and policy consequence.

How do recruiters catch AI-assisted candidates?

Chris Morrow, who runs Digitalent, a Manchester recruitment agency that places talent in the AI sector, said the signs are unmistakable. Apps that listen in on virtual interviews and feed candidates suggested answers in real time are already in circulation, he said — and the people using them rarely advance.

"It's like they are reading off an Autocue," Morrow said. "Whenever it happens to us, we tend to end the interview. It's cheating."

Morrow described two tells. First, candidates stall while waiting for the chatbot to generate a response. Second, their language turns stilted.

"There is a delay," he said. "The candidate tends to use filler phrases like 'That's a good question.' The other tell is that a lot of the language they use is not necessarily natural and is super-formal."

In plain terms: recruiters say real-time AI prompting is detectable now, because latency and unnatural phrasing betray it.

What forms does the misconduct take?

The ISE survey breaks the misuse into distinct categories, ranked by frequency:

• Unauthorised use of AI during interviews and online assessments: 66% of employers reported at least one incident
• Using AI to complete online assessments: 47%
• Using AI to complete online application questions, and using it to write CVs or covering letters: 10%

The pattern suggests candidates deploy AI at every stage of the pipeline, from written applications to live conversation. The interviews category — the hardest place to hide machine assistance — still tops the list, which indicates either confidence in the tools or poor judgment about their detectability.

One gap in the data deserves note: the survey measures employer reports of suspected or confirmed incidents, and 92% combines both. The ISE did not publish a breakdown of how many incidents were confirmed versus merely suspected, leaving open how much of the record figure reflects heightened vigilance rather than rising cheating alone.

What is at stake for candidates?

Stephen Isherwood, chief executive of the ISE, framed the risk as a mismatch problem rather than a punishment problem. Cheating can land candidates in roles they cannot perform.

"If you misrepresent yourself, you could end up doing the wrong job," Isherwood said.

His advice to applicants is unglamorous: research the role and understand it before walking into the interview. "That's what interviewers are looking for," he said.

The point is practical. An interview exists to test fit between candidate and job; AI-generated answers defeat that test on both sides. A candidate who passes with machine assistance inherits a position built for skills they do not have.

Is AI actually shrinking the graduate jobs market?

The survey's more consequential finding may be what it does not show: an AI-driven collapse in graduate hiring.

The ISE's 2026 survey reports that the graduate hiring market stabilised this year after an 8% decline last year, rising 4% in 2026. Employers expect a further 7% rise over the next 12 months.

Beneath that aggregate sits considerable variation:

• 40% of respondents are reducing their graduate intake
• 45% are increasing it
• 15% are maintaining current hiring levels

Isherwood attributed the cuts to macroeconomic conditions, not to AI adoption. "They [companies] are not managing down graduate recruitment because of AI. It's because of economic uncertainty. It's a tough economic environment," he said.

He placed the concern about AI's impact on graduate jobs further out on the timeline — "two or three years down the line," in his words.

That assessment cuts against the widespread expectation that automation would already be eroding entry-level hiring. For now, according to the ISE's data, the economy is the variable moving the numbers, and AI's measurable footprint on the graduate market remains confined to the cheating statistics.

The obvious caveat: this is an employer-side survey of ISE members, not a labour-market study, and expectations of a 7% rise are forecasts from the same employers currently reporting record misconduct. Still, on the ISE's evidence, the immediate AI story in graduate recruitment is integrity, not displacement — and the tools candidates use to cheat are, so far, losing to recruiters who say they can hear the difference.