AI interview policy template: three stances you can copy
Pick a stance on AI in interviews, copy the wording for candidates and interviewers, and decide in advance what a flagged answer triggers. Edit it to fit your process.
An AI interview policy is a short document that says whether candidates may use AI during an interview, what they must disclose, and what interviewers do when they suspect it. This page gives you three stances to choose from, a one-page template you can copy, and the part most policies skip: what happens when an answer looks AI-assisted.
Decide the stance first, then write the rest around it. A policy that bans AI but has no way of noticing it is a sign on a door with no lock.
What should an AI interview policy cover?
A usable policy covers six things: the stance, what candidates are told, accommodations, what interviewers do, how a concern is handled, and who owns the policy. Keep it to one page. If an interviewer can't recall it mid-call, it won't get used.
Most employers are already writing one. In the Institute of Student Employers' 2026 survey, 82% of UK graduate employers said they give candidates guidance on acceptable AI use, up from 55% a year earlier (Resultsense, citing The Guardian). The rest of the 2026 numbers are in our AI interview cheating statistics.
Adapt the tone to your brand, and have whoever handles employment law in your organisation read it before you publish. This is a starting point, not legal advice.
Which stance should you pick?
There are three workable stances. Choose by what the interview is meant to measure.
| Stance | Candidate may | Best for |
|---|---|---|
| Ban during the interview | Prepare with AI beforehand, then use nothing live | Technical screens, sales and case interviews, anything testing unaided thinking |
| Prep only, disclosed | Use AI to research and practise, and say so if asked | Most agency and in-house first rounds |
| Disclosed use | Use named AI tools live, if they say so up front | Roles where working with AI is the job |
Most teams land on the first or second. The third is honest, but it means redesigning your questions around how someone works with a tool, not what they know.
The AI interview policy template
Copy this into a doc, replace the parts in brackets, and keep the wording for the stance you picked. It's written for the first stance, with the changes for the other two marked.
1. Purpose
Interviews at [Company] are how we learn how you think and what you've done. This policy explains when you can use AI tools in our hiring process, and what happens if something looks off.
2. Preparing for your interview
You're welcome to use AI tools to prepare: researching [Company], practising answers, or getting feedback on your CV.
3. During live interviews
Please don't use AI tools, notes from an assistant, or another person to help you answer during a live interview, unless your recruiter tells you otherwise in writing. This covers video calls, phone screens and live technical exercises.
Prep only, disclosed: add "If you used AI to prepare, tell us. It won't count against you."
Disclosed use: replace this section with "You may use [named tools] during the interview. Tell us at the start which ones you'll use, and say when you're using one."
4. Take-home tasks
[Say whether AI is allowed on take-home work, and whether candidates should explain how they used it.]
5. Accommodations
This policy doesn't restrict assistive technology used as a disability accommodation. If you need an adjustment, tell [contact] before your interview and we'll agree it with you.
6. Recording and analysis
Our interviews are recorded and analysed by [tool], which joins the call as a named participant. [Add the notice and consent wording that applies where you hire.]
7. How we handle concerns
If an answer looks AI-assisted, your interviewer may ask follow-up questions. A concern on its own is never a reason to reject you. If we confirm that AI or another person helped you during a live interview without being disclosed, we may end your application.
8. Owner and review
This policy is owned by [role] and reviewed every [six months]. Last updated: [date].
9. For interviewers (internal, don't publish)
Ask follow-ups that need the candidate's own experience. Write down exact lines and times, not impressions. Never accuse on the call. Raise a concern with [hiring manager] after the interview, with your notes.
What should you tell candidates?
Say it in the invite, not at minute one of the call. Telling candidates early is fairer, and makes a concern easier to raise later. Here is wording for the first stance:
- Before the interview: "You're welcome to use AI tools to prepare. During the interview, please don't use AI tools, notes from an assistant, or another person to help you answer. We want to hear your own thinking."
- At the start of the call: "As a reminder, this interview is unaided. We'll ask follow-ups on your answers, so there's no need to be perfect."
- If the call is recorded or analysed: say so, and name the tool. Consent obligations sit with whoever runs the interview, so check what applies where you hire.
For the second stance, swap the middle sentence for "If you used AI to prepare, tell us; it won't count against you." For the third, name the tools allowed and ask candidates to say when they're using one.
What should interviewers do?
Give interviewers three rules and nothing more.
- Ask follow-ups that need the candidate's own experience, such as "What did you personally build?" or "What would you change now?"
- Note exact lines, not impressions. "Answer arrived in full sentences with no pause" is evidence. "Felt scripted" is not.
- Never accuse on the call. Probe, and write it up afterwards.
Our guide to telling whether a candidate is using AI in an interview covers the signals worth noticing and the habits that only look suspicious.
What happens when an answer looks AI-assisted?
Decide this before the first interview. The rule we'd suggest is simple: a concern is a signal to follow up, never a verdict and never grounds to reject on its own.
Write the sequence down: ask a different follow-up, compare the answer with the candidate's other answers, then record what you saw and let the hiring decision rest on the whole interview. Two interviewers agreeing on the evidence is stronger than one interviewer's hunch.
Keep a concern separate from a breach. A concern means an answer is worth probing. A breach means you've confirmed undeclared help during a live interview, for example because the candidate said so or couldn't explain their own answer under follow-ups. Only a breach should trigger the consequence your policy names.
Candit is zero-friction anti-cheat for hiring. It joins your Zoom, Meet, Teams, Slack or Webex interview with nothing for the candidate or interviewer to install, and flags answers worth probing while the call is still on, with the exact line behind each flag. On paid plans, Counter AI suggests a follow-up question the moment an answer is flagged, and the interviewer decides whether to ask it. See how AI Detectability works.
How do you enforce it without lockdown tools?
Lockdown browsers and portals add friction for good candidates and still miss a copilot running on a second device, as we explain in our guide to interview proctoring software. A policy that depends on a screen being watched is easy to route around.
A policy you can enforce rests on three things: the question design, your interviewers' notes, and a tool that reads the conversation instead of the candidate's device. Candit has no access to the candidate's device, which is why nothing needs installing, and it flags AI-assisted answers live on the interviewer's screen wherever the copilot is running. That includes the bullet-point format many copilots now use so candidates can put answers in their own words: the pause while the points appear, points delivered in list order, and written phrasing in a spoken answer.
Frequently asked questions
Should candidates be allowed to use AI in interviews?
It depends on what the interview measures. If it tests unaided thinking, ban live use and say so. If the role is built around AI tools, allow named tools with disclosure.
Is it fair to ban AI in interviews but not in the job?
Yes, if you say why. The interview is a sample of how someone thinks, and the job is a different test. Tell candidates which one they're sitting.
What should happen if a candidate breaks the policy?
Name the consequence in the policy, and apply it only to a confirmed breach, not a suspicion. Employers that publish their policies, such as Greenhouse and Match Group, say the application may be disqualified or withdrawn. A flagged answer on its own should lead to follow-up questions, not a rejection.
Should the policy be public?
Yes. A policy candidates can't see can't be enforced fairly. Put the candidate wording in your invite and on your careers page.
Put the policy to work
A policy only counts if interviewers can follow it in the moment. Candit gives your team evidence during the call, so a flag turns into a follow-up question, not an argument afterwards. If you're setting this standard across a team, book a demo.