Why an Employer Can Interview You With AI and Ban Yours

Why Do the Two Rules Point Different Ways?

The rules differ because the two uses do different things to the measurement. An employer picking an AI interviewer is choosing the instrument your answers get recorded on. AI running live inside your answer changes what that instrument records, which is a different act from choosing it. The best-known published rule draws the same line: it aims at the interview itself, not your preparation.

Amazon is one large employer reported to have put that rule in writing. IT Pro reported in March 2025, relaying internal documents seen by Business Insider, that Amazon's guidelines tell candidates not to use generative AI during an interview unless explicitly permitted ¹. The same report says failure to comply may result in disqualification ¹. Amazon's on-record position, given to CNBC and NBC News that month, is that candidates acknowledge they will not use unauthorized tools ¹.

That wording is discretionary, and nothing in the reporting says how many candidates it has actually disqualified.

The distance between preparing with AI and running it live shows up in the survey data. Prospects at Jisc's Early Careers Survey 2026, with 4,990 self-selected UK responses, found interview practice among the fastest-rising uses of AI while use during the interviews themselves stayed noticeably lower ². Read from the report's chart, that is roughly 47% practising or preparing against about 5% using AI in the interview ².

The guidance on this site for employers draws its line somewhere more specific. Whether to let candidates use AI during the interview, or ban it argues for permitting exactly what the job permits, deciding it per role, and putting the rule in the invitation so every candidate for that role gets the same one.

A rule you only discover mid-interview is a badly written rule.

Where the Complaint Is Fair

The disclosure half of the complaint holds up, and candidates have been surveyed on it. In a 2026 survey by Greenhouse, which sells hiring software, 70% of 1,200 US candidates said they were never clearly told up front, and 21% found out that AI was involved only once the interview had started ³. Those are recollections of what people noticed, not an audit of what any consent screen said.

Employers report the mirror image of that. In the Institute of Student Employers' 2025 survey of UK early-careers employers, using AI during interviews without disclosure or permission was the most commonly reported form of candidate misconduct, at 61% ⁴. That figure counts what employers say they suspected or identified, and no verification step sits behind it ⁴.

Suspicion and evidence are not the same thing.

The same survey found many of those employers are not opposed to AI in principle. Half of them, 49%, said they had no problem with candidates using AI during the recruitment process, and only 10% had banned it or added technical measures to stop it ⁵. But 45% had given applicants no guidance at all on when AI use is appropriate ⁵. That last figure is the one you feel from outside, because whatever the employer thinks, there is nothing to read.

Both halves of a policy belong to the employer, and neither of them is yours to guess at.

Candidates also object to the employer's half on its own. In the same Early Careers Survey 2026, a majority called employer AI use unfair at every stage except initial qualification screening ⁶. The objection ran highest around interviews: read from the chart, 83% said AI use in an online recorded interview was not fair ⁶.

That measures what people think is fair, not whether the tools are biased ⁶.

Employers reading this site get the instruction from the other side. What belongs in an AI hiring policy puts permitted candidate AI use, the tools run on candidates, and candidate notice in one document, signed by named people.

What a Reciprocal Policy Looks Like

A fair version of this rule already exists in writing, and it runs both ways on one page. Anthropic's candidate guidance sets rules for applicants and then states what the company does with AI on its own side: job descriptions, interview questions, candidate communications, hiring metrics, interview transcription and sourcing ⁷. It says it does not train on candidate data and does not let Claude make hiring decisions ⁷.

That is a self-description on a company's own careers page, with no audit and no third party behind it ⁷.

The shape is what makes it usable anyway.

The rules for the candidate and the disclosure of the employer's own use sit in one document, so both halves can be read before agreeing to either. Whether the employer talking to you has published the second half anywhere is a fair thing to find out.

The UK has a regulator's version of the same expectation. The Information Commissioner's Office's November 2024 report on AI recruitment tools says recruiters must tell candidates how AI will process their personal information ⁸. The notice should cover what the AI processes, the logic behind its predictions or outputs, and how personal data is used to train or test it ⁸. Those are recommendations from consensual audits rather than a court ruling, and they apply in the UK ⁸.

A candidate cannot enforce a recommendation directly, though it can be cited in a complaint to the ICO.

Ask What the AI Does and Who Reads It

Ask three questions in the invitation thread, before the slot. What the AI does with the recording, whether a person reviews it before any decision, and what the rule for AI on your side actually is. All three are answerable in a sentence, and asking them is ordinary rather than adversarial.

A longer version of what you can ask an employer about how they use AI in hiring covers who to ask and how to phrase it so it reads as logistics.

On your side the durable move is the boring one. Prepare with AI as much as you like, then answer live in your own voice with no second window open. The reason is not the risk of being caught. A copilot tends to put an audible pause before the answer, and what it hands you is generic in exactly the place a follow-up wants the detail only you have.

What an interview copilot costs you in the follow-up works through that mechanism on the tools people actually buy.

If the invitation's rule is unclear, say what you did rather than guess at what was allowed. A specific, true sentence about how you prepared survives the next question. An answer built to look unaided does not, and the employer who wrote no rule is the one who left you deciding.

Read the evidence >