Stop AI interview cheating before it reaches your shortlist

Interviews now have an invisible extra participant

A new kind of tool listens to a call and suggests answers in real time. One such tool, Cluely, describes itself on its homepage as giving "real-time answers, all while completely undetectable", and says it is "invisible to screen share".

That changes what an interview can tell you. Polished answers to standard questions may come from the candidate, or from a tool reading the conversation. Take-home tasks were already easy to complete with AI; now live conversations are too.

Candidates using AI in interviews is not always dishonest. Many people use AI at work, and some roles reward it. The problem is not knowing which skills belong to the person you are about to hire. That is what employers need to solve.

How HireValid responds to AI interview cheating

No tool can promise to catch every case, and claims that something can detect AI with certainty deserve scepticism. HireValid takes a practical approach in three layers.

Test before you talk. Short skills tests come before the interview, under a time limit, with questions drawn from randomised pools. Scenario questions turn on specific details that generic AI suggestions tend to miss, and visual pattern items are awkward to relay to a text tool quickly.

Record evidence, not verdicts. Integrity settings record browser events, such as leaving the test tab or pasting text, and optional camera checks run only with notice and consent. Those signals feed an Integrity Score for a person to review.

Confirm in the interview. Ask every shortlisted candidate to explain one of their answers. Someone who reasoned through it can explain the trade-off easily; someone relaying a tool's suggestion usually cannot.

Can you detect AI cheating in interviews?

Sometimes, but never reliably on its own. Some signs are worth noticing in a video interview: long pauses before fluent answers, eyes tracking text off screen, answers that are polished but generic, and difficulty with follow-up questions. None of these proves anything; nervous candidates, people thinking in a second language and people with some disabilities can show the same signs.

That is why the most dependable response is not detection but design. Structure the interview so that it asks for specifics: what the candidate did, why, and what they would change. Follow up on details. Combine the interview with a short skills test taken earlier, so you are not relying on one conversation. The integrity overview explains what each signal can and cannot show; the guide to reducing AI assistance covers test design in more depth, and our employer's guide to AI interview cheating tools explains how tools such as Cluely and Interview Coder work.

Interview questions that confirm a candidate's own reasoning

Generic questions invite generic answers, whoever or whatever writes them. Questions about the candidate's own work and test answers are harder to hand to a tool, because the answer depends on what the person actually did and thought.

• "Walk me through how you answered this scenario question. What made you rule out the other options?"
• "Tell me about a time a plan changed halfway through. What did you do first, and what would you do differently?"
• "Here is a slightly different version of the problem. How would your answer change?"

Ask the same questions of every shortlisted candidate and score the answers against a rubric agreed in advance. Our interview templates include structured questions for each role.

A fair response to AI in hiring

The goal is evidence of real ability, not catching people out. Tell candidates clearly which tools are allowed at each stage, and why. Make accommodations easy to request, and do not treat assistive technology as cheating. When a signal appears, ask about it before deciding anything.

Some roles reasonably allow AI in a later work sample, because that is how the job is done. In that case, keep the timed screening tests for independent skills, allow AI in the work sample, and ask candidates to explain their work. Our fairness approach sets out how we think about this balance.