What is AI in recruiting? The uses, the risks and the rules in 2026

AI in recruiting is the use of machine learning and generative AI at any stage of hiring: drafting job ads and outreach, searching and matching candidates, screening applications, scheduling, transcribing and summarizing interviews, and, in some products, scoring or ranking candidates. The closer a tool's output comes to deciding who moves forward, the higher the legal and fairness risk. As of October 2026, New York City requires bias audits and notices for automated employment decision tools, Illinois regulates AI analysis of recorded video interviews, and the EU AI Act classes AI used for recruitment and selection as high-risk.

This page is the overview: the kinds of AI involved, what they are good and bad at, the rules, and where to start. The working policy for using it well is in how to use AI in recruiting responsibly, and the state-by-state law summary is in AI hiring laws by state.

What kinds of AI are used in recruiting?

"AI" covers several different technologies, and they fail in different ways. Knowing which one a product uses tells you what to check.

If a tool sits on the meeting itself, what is an AI note taker explains the capture methods and the consent questions in more detail.

Which uses of AI in recruiting are low risk and which are high risk?

The useful question is not "does it use AI?" but "how directly does its output affect whether a specific person moves forward?" Three broad groups cover most tools:

• Drafting and assisting: the AI helps a person write or organize, and produces nothing about a specific candidate. A job ad draft is the classic example. Risk is low if a person edits it.
• Recording and summarizing: the AI captures or condenses what a candidate said or submitted. Interview notes and resume parsing are here. Risk is moderate: errors and privacy are the issues.
• Evaluating: the AI scores, ranks, filters or recommends candidates. This is where discrimination law, bias audits and notice duties concentrate.

The responsible-use guide turns these groups into tiers with minimum controls for each, so they are not repeated here.

What are the benefits of AI in recruiting?

• Less administration. Drafting, scheduling and note-taking are where recruiters lose most hours that do not involve talking to people.
• More consistent interviews. Tools that build a question guide from the job description, or fill a scorecard as the interview goes, make it easier to ask everyone the same questions and record the same things.
• Better records. A transcript or structured note is easier to check, share and search than handwriting.
• Faster replies to candidates. Drafted updates and scheduling help candidates hear back sooner.

Each benefit depends on someone checking the output. Time saved drafting a submittal is lost, with interest, if the draft goes to a client with a detail the candidate never said.

What are the risks of AI in recruiting?

• Discrimination. A selection tool can produce different outcomes for different groups even when nobody intended it, and federal anti-discrimination law applies whoever or whatever made the selection. The Uniform Guidelines' four-fifths rule of thumb (29 C.F.R. § 1607.4(D)) is the usual first check; see the four-fifths rule.
• Invented or distorted facts. Generative AI writes fluent text that can be wrong, which is most dangerous in candidate summaries and submittals.
• Privacy. Resumes, transcripts and notes are personal data. Pasting them into a tool your organization has not approved can breach your own policy, a client contract or privacy law.
• Candidate trust. Candidates who discover AI they were not told about tend to assume the worst about what it did.
• Over-reliance. A recruiter who sorts by match score and never scrolls has handed the decision to the score.
• AI on the other side of the table. Candidates use AI too, from polishing resumes to, in some cases, getting answers during live interviews. Some apps advertise exactly that: LockedIn AI's site says it listens to your interview and provides real-time answers (as of October 2026). See how to detect AI-generated interview answers and, for fake identities, how to spot candidate identity fraud.

What laws apply to AI in recruiting?

As of October 2026, the main AI-specific rules affecting recruiting include:

Other states, including Colorado and California, have their own rules and timelines, summarized with links in AI hiring laws by state. The EU detail, including what applies already, is in the EU AI Act for recruiting, and the New York City audit itself is explained in the Local Law 144 bias audit guide.

Is every kind of recruiting automation "AI"?

No. A knockout question that rejects anyone without a required license, or an email that goes out when a candidate changes stage, follows a rule a person wrote. AI produces outputs that nobody wrote line by line, from patterns in data or from a language model. The distinction matters less than it seems, though: laws tend to define the tools they cover by what the output does to a hiring decision, and some definitions are broad. Read the definition in the rule that applies to you rather than relying on whether a vendor calls its feature AI.

What should you ask a vendor about its AI?

Marketing pages rarely answer these, so ask in writing and keep the replies:

• What exactly does the AI produce: text for a person to edit, a summary, or a score, rank or recommendation about a candidate?
• Which model or provider does it use, and does that provider receive candidate data?
• Is candidate data used to train any model, and can that be switched off by contract?
• Has the tool been tested for different outcomes by sex, race or ethnicity, and can we see the results and the date?
• Can a recruiter see why a candidate received a particular output, and overrule it?
• What do you recommend we tell candidates, and in which jurisdictions does your product need a notice or audit?

A vendor that cannot answer the first question clearly is not ready for the rest.

How should a recruiting team start using AI?

1. List what you already use. AI often arrives inside tools you already pay for. Check your ATS, sourcing and scheduling tools.
2. Start with low-risk drafting. Job ads, outreach and interview questions, edited by a person. ChatGPT prompts for recruiters has copy-ready prompts with checks.
3. Set one data rule now. No candidate data in tools or accounts your organization has not approved for it.
4. Check every output that names a candidate against the source before it goes anywhere.
5. Tell candidates. AI disclosure templates have wording for careers pages, job posts and the start of a call.
6. Treat evaluation tools as a project, not a purchase. Anything that scores, ranks or filters people needs legal review, bias testing and a named owner before it goes live.