Nearly 4 in 10 Candidates Are Getting AI Help During Interviews. Most Recruiters Cannot Tell.

A study of 19,368 interviews conducted between mid 2025 and early 2026 found that 38.5% of candidates showed signs of using AI assistance during live interviews. The rate tripled in just three months, climbing from 9% to 45% between July and September 2025. It has not come back down since.

If your agency conducted 100 interviews last quarter, the data suggests roughly 40 of those candidates were getting real time help you could not see. And 61% of the candidates who cheated still scored above the passing threshold. They would have moved forward in your process with no detection at all.

This is not a tech industry problem. It is a recruitment industry problem. And it is already changing how the best agencies think about what happens after the interview ends.

What Candidates Are Actually Doing

The tools available to candidates in 2026 go far beyond copying a question into ChatGPT. Dedicated interview cheating applications now run as invisible overlays on the candidate’s screen. They capture the interviewer’s audio, process it through an AI model in real time, and display suggested answers directly on screen. The candidate reads the answer while appearing to look at the camera. From the interviewer’s side, nothing looks unusual. No tabs are switched. No second screen is visible.

Gartner data from late 2025 shows that almost half of all applicants now use AI at some point in the job search process. That includes polishing CVs and preparing answers, which is largely harmless. But 13% of job seekers admit to using chatbots in real time during live interviews. And 6% have admitted to outright interview fraud, including impersonation.

The market for these tools is growing because the economics make sense from the candidate’s perspective. A monthly subscription costs between 20 and 50 euros. The salary difference between getting and not getting the job makes that feel like a riskless bet. And the tools are getting better every month.

Companies like Google, L’Oreal, and Anthropic have already responded by requiring at least one in person interview round. But for recruitment agencies conducting dozens of remote interviews every week, eliminating video interviews is not realistic. The question becomes something different entirely.

Why a Transcript Alone Cannot Catch This

Here is the uncomfortable truth. A polished answer sounds the same whether the candidate said it from experience or read it from an invisible overlay. If all you capture from an interview is a transcript and a summary, you have no way to distinguish between the two.

Experienced recruiters describe the same pattern over and over. The candidate pauses for a moment after every question. Then they deliver an unusually structured, textbook perfect response. But when pushed with a specific follow up question, the depth disappears. The first answer is always strong. The second is vague. The third falls apart completely.

AI tools generate excellent initial responses. They struggle with sustained, contextual conversation. They cannot reconcile what a candidate just said with what is on their CV. They cannot explain why a career decision was made five years ago in a way that connects to the role being discussed. They cannot account for gaps between claimed experience and the depth of the answer.

This is why the recruiters who are catching it consistently are not relying on behavioural cues like eye movement or unusual pauses. Those signals are unreliable. Nervous candidates look exactly like cheating candidates. Instead, the best recruiters are looking at the gap between what was said and what should be true based on the candidate’s actual background.

Context Is What Separates a Good Interview From a Reliable One

A transcript tells you what was said. It does not tell you whether what was said makes sense given who the candidate actually is. That requires context. Specifically, it requires combining the conversation with the candidate’s CV and the job description being recruited for.

When you layer those three data sources together, patterns emerge that are invisible in a transcript alone.

  • A candidate claims deep expertise in stakeholder management, but their CV shows two years of experience in an individual contributor role with no direct reports
  • A candidate delivers a technically perfect answer about a specific methodology, but the job description does not require it and their background suggests no exposure to it
  • A candidate’s stated salary expectation, notice period, and availability do not align with what was discussed in the conversation
  • A candidate provides a detailed answer about leading a team of fifteen, but their previous employer is a three person startup

None of these inconsistencies are visible in a transcript. All of them become obvious when the transcript is combined with the CV and the role requirements. This is what turns raw interview data into a structured candidate report that a recruiter or hiring manager can actually act on.

The report does not just summarise the conversation. It highlights where the candidate’s responses align with the role and where they do not. It surfaces gaps that need follow up. And it does this consistently across every interview, regardless of which recruiter conducted it.

What This Means for Recruitment Agencies Running at Scale

For an individual recruiter, the AI cheating problem is frustrating. For an agency running 50, 100, or 200 interviews a week across a team of recruiters, it is a structural risk. Because it is not just about whether one candidate fooled one recruiter. It is about whether your team’s interview data is reliable enough to make placement decisions on.

When every interview produces a structured report that combines what was said with the candidate’s actual profile and the requirements of the role, the data becomes self correcting. A recruiter who missed something in the moment can see it in the report. A hiring manager receiving the report can evaluate the candidate based on evidence, not on how polished their delivery was. And across the team, the quality of candidate assessment becomes consistent rather than dependent on whoever happened to conduct the interview.

This is also where the conversation type matters. Candidates using AI cheating tools are overwhelmingly doing so on video calls, because that is where the overlay tools work. Phone calls and in person meetings are far harder to game. A candidate sitting across a table from a recruiter cannot read from an invisible screen. A candidate on a phone call cannot paste questions into a chatbot without the recruiter noticing the silence.

Agencies that capture data from all conversation types, not just video, have a more complete and more reliable picture of every candidate. The phone screening call, the in person meeting, and the video interview all contribute to the same structured report. That is three data points instead of one. And the more data points you have, the harder it is for any single interaction to be misleading.

The Interview Is Still the Most Valuable Signal in Recruitment. Protect It.

AI is not going to stop candidates from trying to game the process. The tools will keep getting better. The overlays will keep getting harder to detect. And recruiters will keep facing the same question after every strong interview. Was that really them?

The answer is not to stop doing interviews. The answer is to make the interview data work harder. Combine the transcript with the CV and the job description. Produce a structured report that reveals whether the conversation matches the candidate’s actual profile. And do this for every interview, across every recruiter, across every conversation type.

That is what turns interview data from an opinion into evidence. And evidence is something AI cheating tools cannot fake.

See what a structured candidate report looks like for your agency. Book a call and we will walk you through it using a real interview from your workflow.