How do I write an AI-use policy for candidates in job interviews?
TL;DR: A good AI-use policy for candidates uses a three-tier structure: AI is fine for prep (researching the company, practicing answers), disclosure is required for AI-assisted submitted work (take-homes, portfolios), and AI is prohibited during any live, real-time assessment (interviews, live coding, oral exams). State it before the interview, get written acknowledgment, and back it with proctoring so the rule is enforceable, not just aspirational.
The claim
Employers that publish a clear AI policy get fewer disputes and cleaner disqualifications. Employers that don't get exactly what several companies describe on LinkedIn and Reddit: awkward mid-interview confrontations, no policy to point to, and a rejected candidate who claims they didn't know the rules. A written policy plus enforcement (proctoring) beats either one alone.
The evidence
Greenhouse, one of the largest ATS providers, publishes its own candidate-facing AI guidelines: candidates are told AI is welcome for prep but not for reading AI-produced answers live, and violations can mean disqualification. Multiple state and local laws are moving in this direction too — Illinois's Artificial Intelligence Video Interview Act requires employers to notify candidates before the interview if AI will evaluate them and get written consent. That's a strong signal this is now a category legal counsel expects HR to have covered, not just a nice-to-have.
Comparison: three-tier policy structure
| Tier | Rule | Examples | Disclosure required |
|---|---|---|---|
| Green — prep | AI used before any live assessment | Researching the company, practicing with a chatbot, resume polishing | No |
| Yellow — submitted work | AI used to produce work you turn in | Take-home assignments (if allowed), portfolio pieces, coding tasks where the brief allows tools | Yes, state it openly |
| Red — live assessment | AI answering for the candidate in real time | Live coding help, reading a screen during a video interview, a voice clone or proxy standing in | Prohibited outright |
Step-by-step: writing your policy
- Name the three tiers explicitly. Don't leave "AI use" undefined — candidates genuinely disagree on what counts, so spell out prep vs. submitted work vs. live assistance.
- Send it before the interview, not during. Include the policy in the interview invite so a violation is unambiguous, not a surprise raised mid-call.
- State the consequence. "Undisclosed AI use in a live assessment may result in disqualification" — plain language, no hedging.
- Get written acknowledgment where your jurisdiction requires it. States and cities are starting to mandate this (Illinois's AIVIA is one example); check what applies to your hiring locations.
- Back the policy with detection. A policy without enforcement is just a suggestion. Proctoring — browser lockdown on assessments, an AI Meeting Proctor on live rounds — is what turns "please don't use AI live" into something you can actually catch and document.
- Review annually. AI capabilities move fast; a policy written in 2024 language about "don't use ChatGPT" won't cover voice clones or agentic browsers by default.
FAQ
Should candidates be allowed to use AI to prepare for interviews? Most published policies (including Greenhouse's) say yes — prep is treated the same as reading a book about interview technique. The line is real-time use during the live assessment itself.
Do I legally have to disclose that AI is evaluating candidates? In some jurisdictions, yes. Illinois's Artificial Intelligence Video Interview Act requires notice and written consent before AI evaluates a video interview. Check state and local requirements for every location you hire in.
What should the policy say about take-home assignments? Be explicit: either AI is allowed and must be disclosed, or it's banned entirely for that exercise. Silence is what creates disputes later.
How do I enforce a "no live AI" rule without a proctoring tool? You can rely on interviewer judgment alone, but response-lag and phrasing cues are easy to miss on a bad connection. A tool that monitors gaze, audio, and screen behavior during the call turns a judgment call into documented evidence.
Does a written policy actually reduce disputes? Yes — candidates who are told the rules upfront and asked to acknowledge them have far less room to claim they didn't know, which matters if a disqualification is ever challenged.
By Pinal Dave Last updated: 2026-08-02