Workflow · People and HR

Should AI screen or rank your job candidates? Not yet

Why we say wait, and how to use AI in hiring without letting it decide who gets through.

WaitRules apply to this use that most organisations are not set up to meet.

For: Recruiters, hiring managers and owners of small firms

What the evidence says

Studied directly. Studies look at this task or something very close to it.

AI in screening and interviews: not enough evidence yet (evidence strength 0 of 5)

Too few strong studies to rate. Some results are promising in a specific setting, which is not the same as being safe to use on your applicants.

This workflow itself is our recommended way of working, not a tested method. Studies searched to 2 November 2026.

The rules

Screening is regulated in some places. The EU AI Act treats AI used to recruit, filter or evaluate job applicants as high-risk (Annex III, point 4). New York City bars automated hiring tools unless they've had a bias audit in the past year, the results are public and notices have been given.

EU AI Act, official text · What applies when in the EU · New York City Local Law 144

Checked 5 November 2026. This is not legal advice.

Before you start

This is not legal advice. Hiring law differs by country and sometimes by city. Ask a qualified person what applies to you.
Ask your software supplier whether AI is already ranking or filtering candidates. It may be switched on inside a tool you already use.

What to do instead

Replace the words in square brackets with your own. Stay in the same chat for the next prompt, unless a step says otherwise.

You
Write what the job needs and what you'll assess, before anyone sees an application.

Use AI on the job, not the people: tighten the ad and remove requirements you don't need.

Prompt

Here is a job ad. Point out requirements that aren't needed to do the job, wording that may put off qualified people, and jargon. Flag anything that could discriminate by age, sex, race, disability or another protected trait. Suggest edits. Do not add claims about pay, benefits or culture.

[paste]

Use it to prepare structured interview questions and a scoring guide that you then approve.

Prompt

From this job description, suggest 6 interview questions that test the skills it lists, with a guide to what a weak, fair and strong answer looks like. Don't include questions about age, sex, race, family, pregnancy, health, disability, religion, sexual orientation or nationality. If the job description asks for something unfair or unlawful to test, tell me.

[paste]

You
Read the applications yourself, and score them against your own guide. Keep AI out of who moves forward.
You
Keep a record of why each person did or didn't progress.

How to catch its mistakes

Nothing about a named applicant goes into an AI tool.
Applicants know about any automated step where the law requires it.
If a supplier's tool does rank people, get its audit results and its legal position in writing before relying on it.
Compare pass rates across groups, if you can lawfully do so and have enough applicants for it to mean anything.

When not to use this

Ranking, filtering or rejecting people with AI, at least until you have legal advice, a bias audit and a person who actually reviews each result.
Pasting CVs into a general AI assistant to pick the best ones. Don't do this.

Tools

A general AI assistant, for the job ad and interview questions only. Examples: ChatGPT, Gemini, Microsoft Copilot. We chose these because they are common in workplace software, not because they are the best. We leave out Claude products as examples, because we draft this site with Claude. AI assistants compared has them all, with dated facts. Use whichever your employer has approved. The steps work with any tool of this type. We take no money from AI companies.

If you manage people

Make one person answerable for each hiring decision. Ask HR and IT to list any AI features active in your hiring tools.

Written by AI Adoption's editors as our own recommendation, last checked 5 November 2026. It has not been reviewed by an outside practitioner. Before publishing, we ran its prompts once on made-up samples with a second AI model, also from Anthropic. That is a check on the wording, not a test of whether the workflow works for you. Disclosure: AI Adoption uses AI tools in its research and editorial work, including Claude from Anthropic. No company has any say in this page. Reuse under CC BY 4.0. Spotted an error? Tell us.