AI hiring policy template for your company, ready to copy
A copyable AI hiring policy template with allowed and banned uses, candidate notice language, a rollout plan, and vendor questions for hiring managers.
9 min read
Key takeaways
- An AI hiring policy should name approved tools, list allowed and banned uses, require human review of every decision, and explain how candidates are told.
- Never let an AI tool reject or advance a candidate without a person looking at that candidate's actual materials.
- Tell candidates about AI before the step happens, and give them a clear way to ask for an alternative or an accommodation.
- Check pass-through rates at each AI-assisted step before a new tool goes live and at least once a year after that.
- Rules on automated hiring tools vary by location, so have your legal team review the policy before you publish it.
An AI hiring policy is a short written document that says which AI tools your company uses in hiring and what those tools may and may not decide. It also sets out how candidates are told about AI and protected from unfair decisions. Below is a template you can paste into a doc and fill in, followed by a table of allowed and banned uses, a seven-step rollout plan, and the questions to ask vendors before you approve a tool.
Many hiring teams already use AI somewhere without ever writing it down. A recruiter asks a chatbot to draft a job post. Meanwhile the ATS ranks applicants and the video interview tool summarizes each answer. A policy turns those scattered habits into rules people can follow and explain to a candidate who asks.
What should an AI hiring policy cover?
A useful policy covers six topics in plain language:
- The tools that are approved, and the person who approved each one.
- The tasks AI can do on its own, and the ones that need a person.
- The notice candidates get about AI, and the point in the process when they get it.
- The way a candidate asks for an alternative or an accommodation.
- The candidate data that goes into the tools, plus where it is stored and how long it is kept.
- The checks you run to make sure the tools treat groups of candidates fairly.
Keep it short. Two or three pages that hiring managers read will do more than a long document nobody opens.
AI hiring policy template you can copy
Fill in everything in [brackets] and delete sections that do not apply. Rules about automated hiring tools differ from place to place, so check with your legal team before you publish.
1. Purpose and scope
[Company name] uses AI tools to help recruiters and hiring managers work faster and more consistently. This policy covers every stage of hiring, from writing job posts to making offers. It applies to anyone hiring on our behalf, including contractors and outside recruiting agencies, for roles in [locations].
2. Definitions used in this policy
"AI tool" means any software that generates text, or ranks, scores, filters, summarizes, or recommends candidates using machine learning or a large language model. This includes AI features inside tools we already use, such as our ATS or video interview platform.
"Hiring decision" means moving a candidate forward or rejecting them, including making an offer.
3. Approved tools and who keeps the list
Only tools on the approved list may be used with candidate data. [Owner, for example the head of talent] keeps the list at [location]. To add a tool, send [owner] the vendor's answers to our vendor questionnaire. Employees may not put any candidate information, such as resumes or interview content, into personal accounts on public chatbots.
4. Allowed uses of AI tools
AI tools may be used to:
- Draft job descriptions and candidate emails, which a person reviews before use.
- Draft interview questions, which the hiring manager approves.
- Summarize candidate materials like resumes and interview answers for a human reviewer.
- Suggest scores against a rubric our team wrote, as advice to a reviewer.
- Schedule interviews and send reminders using approved message templates.
5. Prohibited uses of AI tools
AI tools may not be used to:
- Reject or advance a candidate without a person reviewing that candidate.
- Score candidates on facial expressions, voice tone, accent, or appearance.
- Infer or consider age, race, gender, disability, religion, national origin, pregnancy, or any other protected trait.
- Apply criteria that are not tied to the job and written in the role's scorecard.
- Search candidates' personal social media to build a profile.
6. Human review of every decision
A named person makes every hiring decision. That person reads or watches the candidate's actual materials, not only the AI summary or score. If the reviewer disagrees with an AI score, the reviewer's judgment wins, and they note the reason in [ATS field]. No one is penalized for overriding a tool.
7. Candidate notice about AI
We tell candidates when AI is used in their process, in the job post or first email, before that step happens. The notice explains what the tool does and does not do. It also says that a person makes the decision. [Add consent language if it is required where you hire.]
A sample notice could read as follows.
"We use software to help our team review applications and recorded interviews. It can summarize answers and suggest a score against the skills listed in this job post. A member of our hiring team reviews every candidate and makes all decisions. To ask for a different format or an accommodation, reply to this email."
8. Accommodations and alternatives
Any candidate may ask for an accommodation or an alternative to an AI-assisted step, such as a live phone call instead of a recorded interview. Extra time and captions are also common requests. Requests go to [contact] and get a reply within [two business days]. Asking never affects a candidate's chances.
9. Data handling and retention
Only data needed to assess the role goes into AI tools. Candidate data is kept for [period] and then deleted, unless the law requires longer. Vendors may not use our candidate data to train models for other customers without written approval. Candidates can ask what data we hold about them by contacting [contact].
10. Fairness monitoring at each AI-assisted step
Before a new tool goes live, and at least every [six or twelve] months after, [owner] reviews pass-through rates at each AI-assisted step by group, where we hold that data lawfully. A large gap triggers a review of the tool and the rubric it scores against.
11. Vendor requirements for AI tools
Vendors must answer our questionnaire, share any bias testing they have done, tell us before major model changes, and support candidate deletion requests.
12. Ownership and review
[Owner] maintains this policy. It is reviewed every [12 months], and whenever we add a tool or start hiring in a new location. Questions go to [contact]. Last updated [date].
Allowed vs not allowed at a glance
Print this table and keep it next to the policy. It tells a recruiter whether AI is fine for a given task faster than the full text does.
| Task | Allowed? | Condition |
|---|---|---|
| Draft a job post | Yes | A person edits it before posting |
| Write interview questions from a job description | Yes | Hiring manager approves the final set |
| Summarize a recorded interview | Yes | Reviewer still watches the answers |
| Suggest a score on the team's rubric | Yes, as advice | Reviewer makes the final call |
| Send scheduling messages and reminders | Yes | Uses approved templates |
| Auto-reject anyone below a score | No | None |
| Rank on "culture fit" with no written rubric | No | None |
| Analyze facial expressions or tone of voice | No | None |
| Paste resumes into a personal chatbot account | No | None |
The row that does the most work in practice is advice-only scoring, where the tool suggests and a person decides. Look for tools built that way. In DeepHire, for example, AI summaries and scores are based on the recruiter's own rubric and are advice only, so the reviewer still makes the call.
How do you roll out an AI hiring policy?
Step 1: Inventory what you use today
Ask everyone who takes part in hiring, from coordinators to hiring managers, to list each AI tool or feature they touch. That includes AI features built into tools you already pay for, like the ATS or interview software. Job boards count too, and so do your agencies. Expect a few surprises, like a manager who screens resumes with a free chatbot.
Step 2: Sort each use
Put every item into one of three buckets: allowed as is, allowed with conditions, or stop. The table above is a good starting point.
Step 3: Fix the rubric before the tool
AI scoring is only as fair as the criteria you give it. Write a short scorecard for each role with job-related skills and clear rating levels. The hourly candidate scorecard template shows the format for high-volume roles.
Step 4: Get legal review
Some places have rules about notice, consent, bias audits, or recorded interviews. Your legal team should confirm what applies in each location where you hire.
Step 5: Update candidate messages
Put the notice in job posts and in every interview invite, and make the accommodation contact easy to find. If your team needs a refresher on what accommodations look like in practice, see ADA hiring basics for recruiters.
Step 6: Train the people who hire
A 30-minute session is enough. Walk through the table. Then show one example of a reviewer overriding an AI score and where they logged the reason.
Step 7: Book the first fairness check
Put the first review on the calendar now, while the rollout is fresh. Our guide to running an AI hiring bias audit explains what to measure.
What to ask an AI hiring vendor before you approve it
Send these questions with any new tool request. You can drop them into a video interview software RFP or a shorter questionnaire.
- What exactly does the AI do, and what data does it use to do it?
- Can it make or trigger a rejection without a person? Can that be turned off?
- Does it analyze anything beyond the content of answers, such as faces or voices?
- Can we set the scoring criteria, or does the vendor set them?
- What bias testing have you done, and can we see a summary?
- Is our candidate data used to train models for other customers?
- How long is data kept, and how do we delete a candidate's records?
- How will you tell us when the model changes?
A vendor that can't answer these in plain language has given you a good reason to keep looking.
Mistakes that make an AI hiring policy useless
A policy that lives only in a legal folder does nothing. If hiring managers have never seen it, they can't follow it. Send it to them and walk through the table together.
Banning AI outright tends to backfire. People keep using it quietly, and you lose any view of what is happening. Approving a few tools with clear limits works better.
Watch for rubber-stamp review. If reviewers approve every AI recommendation without opening the candidate's materials, the human review exists in name only. Spot-check a handful of decisions each quarter to see whether reviewers actually read or watched what the candidate submitted.
Give the policy one named owner. When ownership is shared, the yearly review is easy to skip, and the approved tool list drifts out of date.
Frequently asked questions
Do we need an AI hiring policy if we only use AI to write job posts?
Yes, but it can be short. Even a one-page policy stops people from pasting candidate data into unapproved tools and sets rules before AI spreads into screening. Writing the rules early is easier than unwinding habits later.
Is it legal to use AI in hiring?
In most places you can use AI tools in hiring, but some places add requirements such as telling candidates or running bias audits. Anti-discrimination rules still apply to decisions a tool influences. Check with your legal team for each location where you hire.
Should candidates be told when AI is used to screen them?
Yes. Tell them before the AI-assisted step and explain what the tool does and that a person makes the decision. Give them a contact for accommodations too, since some places require this notice by law.
Can an AI tool automatically reject job applicants?
A sound policy says no. Let AI summarize and suggest scores, but require a person to review each candidate before any rejection or advancement. That protects candidates and gives you a clear record of who decided and why.
How often should an AI hiring policy be reviewed?
Review it at least once a year, and also when you add a new AI tool or start hiring in a new location. A vendor's model change is another good trigger. Pair each review with a check of pass-through rates at AI-assisted steps.