AI use policy for interviews: a template for interviewers and candidates

An AI use policy for interviews is a short document that says which AI tools interviewers may use, what each tool may do (transcribe, summarize, suggest, score), how candidates are told and asked for consent, which decisions always stay with a person, where AI output is kept and for how long, and what candidates themselves may use. Below is a copy-ready template with a section for each, a filled example of the approved tools table, and notes on the laws that change the wording.

This page is the policy document. For the method behind it (sorting AI uses by risk, checking output and keeping an AI register), see how to use AI in recruiting responsibly. For the candidate-facing part in more depth, see candidates using AI in interviews.

What should an AI use policy for interviews cover?

1. Scope. Which interviews and calls, and who the policy applies to: recruiters, hiring managers, panel members, agency partners.
2. Approved tools. A named list, with what each tool may and may not do.
3. Consent and notice. What candidates are told in writing, what is said at the start of the call, and what happens if they decline.
4. Human decisions. Which decisions a person makes and records, and what AI output may inform them.
5. Checking AI output. What interviewers verify before notes, scores or submittals are saved or shared.
6. Data. Where transcripts and notes are stored, who can access them, how long they are kept and how deletion requests are handled.
7. Prohibited uses. The things nobody does, such as personal AI accounts for candidate data.
8. Candidate use of AI. What candidates may use, and how interviewers respond to suspected live help.
9. Accommodations. How a candidate opts out of AI tools or asks for an alternative.
10. Ownership and review. Who owns the policy, how to propose a new tool, and the review date.

The AI use policy template

Copy it into your own document and replace everything in square brackets.

AI USE IN INTERVIEWS POLICY
[Company name]          Version [x]          Effective [date]
Owner: [name, role]     Next review: [date]

1. SCOPE
This policy applies to everyone who interviews candidates for
[company / our clients], including recruiters, hiring managers, panel
interviewers and [agency partners]. It covers phone screens, video and
in-person interviews, intake calls and debriefs.

2. APPROVED TOOLS
Only the tools listed in Appendix A may be used in interviews or on
candidate data. Each tool may be used only for the purposes listed for
it. Personal or free accounts of any AI tool must not be used with
candidate information. To propose a new tool, contact [owner].

3. CONSENT AND NOTICE
3.1 Candidates are told in the interview invitation that [the interview
    may be transcribed / AI tools are used as described in Appendix A].
3.2 At the start of every captured call the interviewer says:
    "[approved consent line]"
3.3 If a candidate declines, the interviewer stops capture, takes notes
    by hand and records the decline. Declining does not affect the
    candidate's assessment.
3.4 Where a law requires specific notice, consent or timing, the
    jurisdiction notes in Appendix B apply in addition to this section.

4. HUMAN DECISIONS
4.1 Decisions to advance, reject, submit to a client or offer are made
    by a named person and recorded with a job-related reason.
4.2 AI tools may [transcribe, summarize, draft notes and suggest
    questions]. AI tools may not [reject candidates, rank candidates for
    a decision, or send assessments to clients without review].
4.3 Interviewers rate candidates against the scorecard using evidence
    from the interview. An AI-generated rating, if any, is a draft that
    the interviewer confirms or changes.

5. CHECKING AI OUTPUT
Before saving or sharing AI-generated notes, interviewers check:
  a) pay, notice period, start date, dates and employer names against
     the transcript;
  b) every quotation appears in the transcript word for word;
  c) nothing refers to protected characteristics, health, family or
     other personal matters not related to the job;
  d) the summary matches the candidate's actual words ("supported" is
     not "led").

6. DATA
6.1 Transcripts and AI notes are stored in [system]. Access is limited
    to [the interviewer, the recruiter on the requisition, ...].
6.2 Transcripts are kept for [period]; notes and scorecards for
    [period], per the candidate data retention policy.
6.3 Requests to access or delete a candidate's data go to [contact]
    and are answered within [period].
6.4 Candidate data is not used to train AI models, and approved vendors
    are bound by a data processing agreement.

7. PROHIBITED USES
  - Pasting resumes, transcripts or notes into unapproved AI tools
  - Capturing a call without the notice in section 3
  - Using AI to infer age, health, ethnicity, emotion or personality
    from voice, video or text
  - Using transcripts to monitor interviewers without telling them

8. CANDIDATE USE OF AI
Candidates may use AI to prepare and to draft application materials.
During live interviews they are asked to answer without live help
from AI or other people. Interviewers who suspect live help ask
follow-up questions, score only evidence the candidate can explain,
and inform the recruiter. No candidate is rejected on suspicion alone.

9. ALTERNATIVES AND ACCOMMODATIONS
A candidate may ask for an interview without AI tools, or for an
accommodation, by contacting [contact]. Requests are handled through
[accommodation process] and do not affect the assessment.

10. TRAINING, BREACHES AND REVIEW
Interviewers complete [training] before using an approved tool.
Breaches are reported to [owner]. This policy is reviewed every
[12] months and whenever a tool or applicable law changes.

APPENDIX A: APPROVED TOOLS (see example table)
APPENDIX B: JURISDICTION NOTES (see table of laws)

How do you fill in the approved tools table?

Appendix A is where the policy becomes specific. One row per tool, written so a new interviewer can tell in ten seconds whether a use is allowed. An invented, filled example:

Be honest in the "May not" column. If a feature you have switched off would put the tool in a stricter legal category, say it is switched off and who checks that it stays that way. The AI register in how to use AI in recruiting responsibly is a good source for these rows.

Which laws change the wording?

Appendix B is where local rules go. The table shows which section of the template each kind of rule touches. The rules themselves, dated and linked to the statutes, are summarized in AI hiring laws by state, which links on to a deeper guide for each one.

If your interviewers use a note taker, read NYC Local Law 144 and AI note takers for where the line sits between transcription and evaluation. Retention periods for section 6 are covered in how long to keep interview notes.

How do you turn the policy into candidate-facing text?

Candidates need two or three sentences, not ten sections. Lift them from sections 3, 8 and 9:

How we use AI in interviews
We [transcribe interviews to keep accurate notes]. Our interviewers
read every note against the transcript and make all decisions
themselves. You are welcome to use AI to prepare; during the interview
we ask for your own answers. To ask for an interview without these
tools, or for an accommodation, contact [name, email].

The AI disclosure to candidates templates have careers-page, job-posting and spoken versions, and the interview recording consent script has the consent lines for section 3.2.

How do you roll the policy out?

1. Inventory first. List every AI feature interviewers actually use, including ones built into tools you already pay for.
2. Draft with the people who interview. A recruiter and a hiring manager will spot the rules that cannot work on a real call.
3. Legal review of sections 3, 4 and Appendix B for the places you hire.
4. Publish with a 20-minute briefing, including saying the consent line out loud.
5. Check after a month. Pick ten interviews and confirm notice was given, notes were checked and data went where section 6 says.

If the policy arrives alongside a new note taker, combine this with the rollout guide so the pilot tests the policy as well as the tool.

Common mistakes in AI interview policies

• Banning "AI" in general. Interviewers keep using the features built into their calendar and video tools, and the policy is broken on day one. Name the tools.
• Promising what is not true. "A human reviews every application" when a filter removes most of them is worse than saying nothing.
• No answer for "no". Every consent routine needs a clear path for the candidate who declines.
• No owner. Without a name and a review date, the policy describes last year's tools.
• Candidate rules nobody tells candidates. A rule on candidate AI use only works if it is in the invitation.