Job search · Updated 8 September 2026

Can You Use AI in a Job Interview? What Employers Actually Allow

There is no industry rule here, and the advice you will find online is mostly written by companies selling either interview-cheating tools or interview-cheating detectors. The employers themselves have published enough to answer the question properly.

It depends on the employer, and employers actively disagree. Amazon told its recruiters that candidates using GenAI in an interview may be disqualified; Goldman Sachs told students it prohibits any external source including ChatGPT; Canva expects engineering candidates to use AI coding assistants and grades how well they direct them. So there is no general answer to memorise — there is a habit to build, which is finding the written policy before every interview and following it exactly. The one rule that holds everywhere: a hidden tool feeding you answers in real time is deception at any company, including the ones that welcome AI.

Four employers, four different answers

These are not edge cases. They are large, well-known employers that graduates apply to, and they land in completely different places.

Amazon. Guidance issued to internal recruiters, reported from documents seen by Business Insider, reads: "To ensure a fair and transparent recruitment process, please do not use GenAI tools during your interview unless explicitly permitted," followed by "Failure to adhere to these guidelines may result in disqualification from the recruitment process." Note the qualifier — unless explicitly permitted. Even the strictest policies are written as defaults, not absolutes.

Goldman Sachs. In an email to students, reported by Fortune in June 2025: "Goldman Sachs prohibits the use of any external sources, including ChatGPT or Google search engine, during the interview process." Spokesperson Jennifer Zuccarelli gave the reason as "We want to hear from our applicants in their own voice." This one is not really about AI — it is a no-outside-help rule that happens to name a chatbot, which is why a quick search during a video call breaks it too.

Anthropic publishes a candidate AI guidance page that splits the process by stage, and it is the clearest published example of how to think about this. On applications, using Claude is encouraged with a condition: "Please create your first draft yourself, then use Claude to refine it," because "We want to see your real experience, but Claude can polish how you communicate about your work." On take-home assessments: "Complete these without Claude unless we indicate otherwise," to "assess your unique skills and strengths." On live interviews: "This is all you — no AI assistance unless we indicate otherwise," because "We're curious to see how you think through problems in real time."

There is a lesson buried in that one. In June 2025, Fortune reported Anthropic telling candidates it wanted "to understand your personal interest in Anthropic without mediation through an AI system." The current published guidance actively invites you to use Claude on your application. Same company, opposite instruction, roughly a year apart. If you prepare from a news article or a listicle rather than the employer own page, you can follow a real quote straight into breaking a real rule.

Canva went the other way. Its engineering blog post, titled "Yes, You Can Use AI in Our Interviews. In fact, we insist" and published in June 2025, describes backend, machine learning and frontend engineering candidates being expected to use tools like Copilot, Cursor and Claude during technical interviews, replacing the computer science fundamentals screening for backend and frontend roles. Canva says it "now inform[s] candidates ahead of time that they'll be expected to use AI tools, and we highly recommend they practice with these tools before their interview."

What Canva assesses is the useful part, because it is what any AI-allowed interview is really testing: whether candidates "understand when and how to leverage AI effectively," how well they "break down complex, ambiguous requirements," and whether they "can identify and fix issues in AI-generated code." Accepting the model output is the failure mode. Catching its mistakes is the assessment.

The line that holds at every company

Across policies that otherwise contradict each other, one thing is consistent: concealment is the problem, not the tool. Canva hands candidates an AI assistant and watches them use it. Nobody, Canva included, accepts a hidden overlay quietly generating answers you then present as your own reasoning. That category of tool is built to be undetectable, and a tool whose main feature is that the other person cannot tell is not a grey area.

The practical test is one question: would I be comfortable sharing my screen right now? If the honest answer is no, you have your answer, and you should assume the employer would reach the same conclusion. This is the same distinction our guide to whether it is cheating to use AI for schoolwork draws for coursework: disclosed assistance is help, concealed substitution is misconduct.

Employers are also primed to suspect it. Checkr surveyed 1,000 US business leaders through Pollfish in April 2026 and published the results that June: 49% said they had extended an offer to a candidate who misrepresented their qualifications using AI tools, and another 26% suspected it but could not confirm. You are interviewing in a market where the person opposite you has probably been burned, which means honest candidates now get scrutiny they did not use to get.

What employers changed to check

Detection software is not the main story, because it does not work reliably. Employers changed the format instead. Reporting in September 2026 by Allwork.space documents companies asking candidates to remove virtual backgrounds, pan their camera around the room or wave a hand in front of their face; verifying claimed location by IP address; requiring live screen sharing; replacing take-home tasks with live written exercises; flagging signals such as browser tab switching, with humans rather than an algorithm reviewing the result; and bringing final rounds back in person onto company equipment.

None of that should worry an honest candidate, but two things follow from it.

• Assume the room is part of the interview. A second monitor angled out of frame, a phone propped beside the laptop, or a browser full of tabs can read badly even when it is innocent. Close what you do not need before you join.
• Practise unaided, not just prepared. Amazon guidance for interviewers reportedly flagged behavioural signals including typing while a question is being asked, appearing to read rather than respond, and wandering eyes. A candidate reading from carefully prepared notes off-screen can look exactly like a candidate reading from a chatbot. Rehearsing until you can speak without the notes protects you from being wrongly suspected, as much as it improves the answer.

Find the policy before you interview

Check three places, in this order.

1. The careers site. Search it for "AI" — a growing number of employers publish a dedicated candidate AI page, and that page overrides anything a news article said.
2. The invitation email and assessment instructions. The rule is usually one line near the logistics, and it is the most commonly missed source because it does not look like a policy.
3. The recruiter, in writing. If the first two are silent, ask. It is not a red flag, and a written reply protects you if it is ever questioned later.

A short message that gets a straight answer without sounding like you are angling for permission:

Hi [name],

Looking forward to the [stage] on [date]. One quick question so I prepare
the right way: what is your policy on using AI tools during this stage?

I want to make sure I follow it exactly - happy either way, I just do not
want to assume.

Thanks,
[your name]

If the answer is that AI is allowed or expected, ask one follow-up: which tools, and whether you should narrate what you are doing as you go. In an AI-allowed technical round, the narration is the assessment.

Three prompts that prepare you without breaking any rule

Preparation is not the interview, and no policy above restricts how you rehearse. These are built so that what you gain survives the tool being switched off.

1. The unaided-recall drill. The point is to stop you writing scripts you will not have in the room.

You are running a mock interview for a [role] role. Ask me ONE question at
a time and wait for my spoken answer before the next.

Rules:
- Do not write me an example answer, ever. Not even if I ask.
- After each answer, give me only: one thing that landed, one thing that
  was vague, and the follow-up a real interviewer would have asked.
- If I give you a rehearsed-sounding answer, say so.

Start with a question about [topic]. Ask me eight in total, then tell me
which two answers I should rebuild before the real thing.

2. The AI-allowed round rehearsal. For a Canva-style interview where you are expected to use AI and judged on how you direct it. What is being graded is your review of the output, so practise reviewing.

Write me a solution to this problem: [paste the problem].

Then stop. Do not explain it and do not tell me whether it is correct.

I am going to review your code out loud as if an interviewer is watching:
what it does, where I think it breaks, what I would test, what I would
change before shipping it.

After I finish, tell me only what I MISSED - bugs, edge cases or bad
assumptions I did not name. Do not praise the parts I got right.

3. The policy-reading check. Useful when the instructions are long and the rule is buried, and it doubles as a way to catch a condition you skimmed past.

Here are the instructions I was sent for a job assessment:
[paste the full text]

Answer only from this text. Do not use outside knowledge about the company.

1. Quote any sentence that mentions AI, external tools, or outside help.
2. Say what is explicitly allowed, explicitly forbidden, and simply not
   addressed - keep those three separate.
3. List the questions I should ask the recruiter to close the gaps.

If the text says nothing about AI, say exactly that. Do not guess what the
policy probably is.

Note the last instruction on prompt 3. A model asked about a company AI policy will confidently invent a plausible one, which is the single most dangerous hallucination in this whole area — our AI hallucination checklist covers the general habit. Never let a chatbot tell you what an employer allows. Only the employer can.

Where this fits in the rest of the process

• The application. Screening happens first and the rules there are different again — see getting your resume past AI screening and tailoring a resume and cover letter with AI.
• Recorded one-way screens. A different format with its own rules and its own candidate rights: preparing for a one-way AI video interview.
• Behavioural rounds. Almost always AI-free by default, and won on story quality. Build the bank in advance with our STAR method practice guide.
• Technical rounds. Whether AI is allowed or banned, the underlying skill is the same — our technical coding interview prep guide builds it.
• What you paste in. Job descriptions are public; your draft answers, referee names and any internal document an employer sends you are not. Most chatbots retain conversations by default, so the privacy checklist applies here too.

FAQ

Can you use AI during a job interview?

It depends entirely on the employer, and employers do not agree. Amazon told recruiters that candidates should not use GenAI tools unless explicitly permitted and may be disqualified; Goldman Sachs told students it prohibits any external source including ChatGPT; Canva expects engineering candidates to use AI assistants in technical interviews. There is no shared standard, so find the written policy before you interview and follow it exactly.

Is using AI in an interview cheating?

The test is not whether AI was involved but whether the employer knows. Using a tool the employer invited you to use is not cheating. Running a hidden overlay that feeds you answers you present as your own thinking is deception at any company, because it is built to be concealed. Quick check: if you would not be willing to share your screen at that moment, you are on the wrong side of the line.

Can employers tell if you used AI in an interview?

There is no reliable detector, so employers changed the format instead. Reported 2026 tactics include turning off virtual backgrounds, camera pans, hand-wave checks, IP location verification, screen sharing, live written exercises replacing take-homes, human-reviewed browser tab-switch monitoring, and in-person finals on company equipment. Amazon guidance for interviewers also flagged signs like typing while being asked a question or appearing to read rather than respond.

Can I use AI to prepare for an interview?

Yes — preparation is not the interview, and none of the policies here restrict how you rehearse. Mock questions, pressure-testing your stories, company research and drilling out loud are ordinary preparation. The constraint is that what you build has to work with the tool switched off, so practise producing answers unaided rather than memorising scripts you will not have in the room.

Bottom line

Stop looking for the rule and start looking for their rule. The employers that have published something are clear, they are easy to find, and they contradict each other enough that guessing is a genuine risk — Anthropic alone reversed its own application-stage advice inside a year. Read the careers site, read the invitation email, and if both are silent, ask the recruiter in writing. Then prepare hard with AI and walk into the room able to do it without.