Why Some Companies Are Banning AI in Interviews — and Others Are Requiring It

If your last two interview invitations felt like they came from different planets, you weren't imagining it. One recruiter tells you flatly not to touch AI during the interview. The next hands you a laptop and asks you to use it. Neither is being difficult — they're each trying to solve the same problem in opposite ways, and knowing which camp you've landed in changes how you should prepare.

What actually changed this year

An analysis of over 19,000 live AI-powered interviews, run by the assessment platform Fabric between July 2025 and January 2026, found that 38.5% of candidates were flagged for AI-assisted cheating behaviour — reading from an overlay, pausing mid-sentence to let a tool generate an answer, that kind of thing. The flagged rate climbed from around 9% in July to 45% by September and has stayed high since. In software engineering interviews specifically, it was closer to 48%.

That's the backdrop. It's why, over the last year, in-person interview requests reportedly rose from about 5% of roles in 2024 to 30% in 2025 — and why the policies candidates are running into now look so different from company to company.

The companies banning it

Amazon and Google reportedly tell candidates plainly not to use AI during interviews, with disqualification on the table for anyone caught. Anthropic bans it outright. The logic from the interviewer's side: live conversation is hard to fake in real time. A follow-up question ("what did you personally do next?") exposes an AI-fed answer fast, because there's no real memory underneath it to draw on.

If you're walking into one of these, the preparation that matters most is the preparation you do before you sit down — not a tool running quietly in the background. That's the whole premise behind rehearsing out loud with a STAR-scored practice session: by the time you're in the room, the example is yours, not a screen's.

The companies building it in

Canva has reportedly gone the other way, restructuring its technical interviews to require candidates to use AI, on the basis that it reflects how the job actually gets done. What's being scored isn't whether you can code without help — it's how you direct the tool, clarify ambiguous requirements, and make the judgement calls a tool can't make for you. Meta is reportedly testing something similar for coding interviews.

A few employers sit in the middle: EY reportedly allows AI for interview preparation but not live in the room, and L'Oréal has leaned back toward face-to-face interviews after noticing that answers polished by AI were starting to sound suspiciously uniform across candidates.

What a redesigned process can look like

One version worth knowing about, used by the recruitment firm QBS Global: a single role-specific live question, a structured 15–20 minute conversation, a short paid work sample (two to four hours), then a live walkthrough of that work — where the interviewer deliberately changes a requirement partway through, to see how you adapt on the spot. Candidates are scored against a rubric of three to five competencies, by more than one scorer.

The common thread across every version of this, whichever side of the AI line a company sits on: the moment that decides the interview is increasingly a live, unscripted one. A tool can help you prepare for that moment. It can't stand in it for you.

Before and after: "walk me through your approach"

Weak (AI-fed, and it shows under follow-up): "I leveraged a cross-functional approach, aligning stakeholders around a shared framework to drive the project toward our key objectives." Follow-up: "Which stakeholder pushed back hardest, and what did you say to them?" — long pause, no real answer.

Strong (prepared, not fed): "I started by mapping who'd be affected — that was five teams, and two of them had conflicting deadlines. I went to the engineering lead first because she had the most to lose, walked her through the trade-off, and we agreed on a staged rollout instead of a single cutover." Follow-up lands easily, because there's a real decision underneath it.

That gap is the whole story. AI can help you find the words. It can't hand you the five-teams, two-conflicting-deadlines detail — only a real example can, which is why rehearsing real examples out loud, ahead of time, matters more now than it did two years ago.

How to find out which policy you're walking into

Ask, plainly, before the interview: "Is AI assistance allowed during this round?" Most recruiters would rather answer that upfront than have it become an issue mid-interview. If the job posting or portal mentions AI tools, screen recording, or integrity checks, take that as your answer. When in doubt, prepare as if it's banned — that preparation serves you either way.

For the ethics side of this — where the line sits, and how panels actually notice AI-fed answers — see our existing article "Using AI in Job Interviews: What's Fair, What's Not, and Can Employers Tell?". And if your next interview is a recorded one-way screen, the one-way video interview guide covers the setup questions this piece doesn't.

Quick questions

Do I have to ask whether AI is allowed, or should I assume it isn't?

Ask if you're unsure — it's a normal question, not a red flag. Absent an answer, prepare as if it's banned.

If a company requires AI use, does that mean coding skill doesn't matter?

No — it means the skill being scored shifts from typing the solution to directing the tool and catching what it gets wrong.

Is in-person interviewing coming back for good?

It's rising sharply for roles where live judgement is hard to assess remotely. Expect it to stay elevated for technical and senior roles even as video interviewing remains the norm for first rounds.