How to Use AI in Your ATS

Which AI features to switch on first, how to test matching on your own roles, and where a recruiter should stay in the loop

Four recruiters gather around a desktop screen in a bright office, reviewing applicant rankings together before deciding which AI features in their ATS to trust

Turning on AI in your ATS (applicant tracking system) is often no longer a separate decision. Vendors add resume parsing, applicant ranking, and drafting tools through routine updates, so some may have gone live before your team agreed on how to use them.

An AI applicant tracking system is an ATS that uses AI to read candidate information, write recruiting text, and sort or filter applicants. The sensible way to adopt those features is by risk level, starting with drafting and search, which a recruiter reviews before anything goes out. Ranking and auto-reject can wait until you've tested them on roles you know, because they decide which applicants a recruiter ever sees.

Key takeaways

  • Group the AI features in your ATS by who makes the final call, since features a recruiter reviews carry less risk than filters.
  • Start with drafting, search, and summaries, then test matching on a recently filled role where you know which applicants were strong.
  • Keep every applicant visible after ranking, review a sample of auto-rejected applicants, and retest both after vendor updates.
  • Notice, audit, and record-keeping rules depend on where the role is based, including New York City, Illinois, and California.

What AI in your ATS can do

ATS AI features fall into three broad groups: reading candidate information, writing text, and sorting or filtering applicants. SHRM's 2026 research of 1,908 in-house HR professionals found that 27% of organizations use AI in talent acquisition, more than any other HR area.

Features that read candidate information

Resume parsing pulls names, job titles, dates, and skills from a resume into profile fields. Database search builds on those fields, and newer versions accept a plain-language request instead of a Boolean string. A parsing error on an unusual layout can hide a candidate from later searches, so test resume parsing on a few resumes with tables or columns.

Features that write text

Writing features run on large language models (LLMs), which generate text from a prompt. Inside an ATS, they draft job descriptions, outreach, and rejection emails, and they summarize interview notes or recorded calls. SHRM's 2024 survey of 2,366 U.S. HR respondents found that 65% of those using AI for recruiting used it to help write job descriptions.

Features that sort and filter applicants

Matching gives each applicant a match score, a rating of how closely the profile fits the stated requirements, and ranking orders the pool by that score. Knockout questions and auto-reject rules remove applicants who miss a set criterion, and automated workflows or agents move candidates between stages on their own. Vendors build scoring differently, so it helps to know how AI in hiring reaches a score before relying on one.

Chatbots can sit in either camp. A recruiting chatbot that books interviews is a convenience. One that asks qualifying questions is screening applicants and deserves the same scrutiny as a knockout rule.

Where AI in your ATS decides who gets seen

AI in your ATS carries the most risk where it decides which applicants a recruiter reviews, and the least where a person reads its output first. A weak job description gets caught in editing. A rule may rank an applicant too low or remove them entirely, and that miss is hard to spot.

FeatureWhat it doesWho decidesWhere to start
Resume parsingPulls resume details into profile fieldsRecruiter reviews the profileSpot-check unusual resume formats
Search across your databaseFinds past candidates from a plain-language requestRecruiter picks who to contactCompare with Boolean results
Drafting job descriptions and messagesWrites first drafts of posts and outreachRecruiter edits before sendingSwitch on early with a review rule
Note and interview summariesCondenses notes and recorded callsInterviewer checks against their own notesUse once interviewers confirm accuracy
Chatbots and schedulingAnswers questions and books interviewsDepends on any screening questionsStart with scheduling only
Matching and rankingScores and orders applicantsRecruiter if every applicant stays visibleTest on a filled role first
Knockout questions and auto-rejectRemoves applicants who miss a set criterionThe system unless a person reviewsAudit current rules before adding more
Automated workflows and agentsMoves candidates between stagesThe system within your rulesAdd last with activity logging on

Review points matter partly because people respond differently to AI suggestions. In a University of Washington study, 528 U.S. participants showed little bias unaided but mirrored AI recommendations, following severely biased picks around 90% of the time. It was a lab study, not recruiters at work, but it's a fair reason to check low-ranked applicants as well as the top.

A rollout sequence for AI in your ATS

The rollout below follows the table from top to bottom, with a check before each step up in risk. That way the team learns how the system behaves before anything filters applicants.

Step 1: An inventory of what's already switched on

The first job is a list of every AI setting that's active now, including any vendor update turned on by default. Knockout questions and auto-reject rules deserve the closest look, since you can copy a rule written for one search into later roles. Note who owns each setting, because the person who switched it on may have moved on.

While you build the list, ask your vendor which fields feed a match score and whether a recruiter can see why one applicant ranked above another. It also helps to know whether the model announces updates and whether it logs auto-rejections.

Step 2: Drafting, search, and summaries first

These make good first features because a recruiter sees and edits every output before it reaches a candidate or hiring manager. A team rule helps: no AI-drafted post or message goes out unread. Whoever edits it should watch for requirements the model added that the role doesn't need. For summaries, ask interviewers to compare the first few against their own notes before the team relies on them in debriefs.

Step 3: Check matching on a filled role before you trust it

Matching is clearest to judge on a search where you already know the answer. Consider a role you filled last quarter, where you know who made the shortlist, who interviewed well, and who got the offer. Run the AI ranking on that applicant pool and see where those people land, then look at anyone you ranked high but passed on.

If strong applicants sit near the bottom, look at what the system weighted. It may have penalized a different job title for the same work, a career gap, or a skill described differently. Repeat the test with different roles, since a ranking that suits one job family may misread another. The same method works for AI resume screening tools outside the ATS.

See why each candidate matches before you shortlist

Book a demo

Step 4: Review points for ranking and auto-reject

Once matching holds up on your test roles, decide where a person reviews its output before it affects a candidate. One option lets ranking order the list while keeping every applicant visible, so a recruiter can scroll past the top results. Auto-reject rules are safest when limited to objective requirements, such as a required license. For the first few months, have someone review a sample of rejected applicants each week to catch rules that filter more broadly than intended.

Step 5: A record of settings and candidate notices

The last step is a short written record of which features are on, what each rule does, who approved it, and when it was last tested. Update it whenever a setting changes, since it's the basis for answering screening questions from hiring managers and for the notices some jurisdictions require.

How to tell whether the AI is working on your roles

The clearest ongoing check is to regularly compare AI shortlists with recruiter shortlists for the same roles. When the two lists diverge, find out why before deciding which is right. Sometimes the system surfaces people a recruiter overlooked, and sometimes it misses context a recruiter picks up from one line on a resume.

Candidate outcomes belong in the same review. Track application completion and drop-off after you add a chatbot or knockout questions, and read the feedback candidates send about the process. If completions fall right after a new screening question goes live, start with that question.

If the built-in features still fall short after a few rounds, AI recruiting software built for one task can go deeper on search or ranking. It needs an ATS integration to read and write candidate records, which adds setup work.

What the rules say about AI in hiring

A handful of places now regulate automated hiring tools. New York City's Local Law 144 bars employers and employment agencies from using an automated employment decision tool without a bias audit within the prior year. Under the city's rules, the audit information must be public, and candidates must receive notice.

Illinois amended its Human Rights Act through Public Act 103-0804, effective January 1, 2026, to bar AI use that has a discriminatory effect on protected classes. The amended act also bars zip codes as a proxy for protected classes and requires employers to notify employees when they use AI. California's Civil Rights Council regulations, effective October 1, 2025, require covered employers to keep employment records, including automated-decision data, for at least four years.

Coverage details vary, so check the rules for each role's location with legal counsel before you change notices or retention settings.

What changes for agency recruiters

An agency ATS holds candidates who may be submitted to several clients. A match score built around one client's job says little about how that candidate fits another client's role, so review rankings search by search. Clients may also ask how you screened candidates, and the settings record from Step 5 answers that.

In-house teams own the whole process, including notices and audits. Some of the rules above cover both employment agencies and employers, so agree with each client who handles those duties.

Before you switch on more AI, check what already filters applicants

A useful next step is to open the auto-reject and knockout rules on your current roles and confirm each still matches what the job requires. Once those are in order, adding ranking or automation becomes a matter of testing it on a role you know.

Search your candidates in plain language

Book a demo

Frequently asked questions

Does ATS software detect AI?

Whether a system flags AI-written resumes depends on the product, so check your vendor's documentation. Either way, the wording of a resume says little about whether a candidate can do the work. Screening calls, work samples, and interviews remain the place to assess that.

What is an AI recruitment system?

An AI recruitment system is a broader term than an AI-powered ATS. It covers any recruiting software that uses AI to source, screen, schedule, or communicate, whether built into an ATS or connected to one. When you compare one, ask which of those tasks its AI handles.

Should you switch to a new AI ATS to get better features?

Switching systems is a large project, so test the AI already in your ATS first and note where it falls short. If the gaps are in core tracking and reporting, comparing the top applicant tracking systems makes sense. If the gap is one task, a dedicated tool alongside your current system may be the smaller change.

About the author

Mihailo Bozic
Mihailo Bozic

Founder & CEO @ Rotto

Founder & CEO of Rotto, building tools that help tech recruiters source better candidates, faster.

Related articles

Candidate Engagement: How to Keep Candidates From Going Cold
Recruiter Playbooks

Candidate Engagement: How to Keep Candidates From Going Cold

Candidate engagement explained: what it is, why candidates drop out, and candidate engagement strategies for each stage from first contact to start date.

By Mihailo Bozic
9-Box Grid: What It Is, How to Run One, and What It Means for Hiring
Recruiter Playbooks

9-Box Grid: What It Is, How to Run One, and What It Means for Hiring

Learn what a 9-box grid is, what each box means, how to run a fair 9-box talent review, and how to turn the gaps it shows into a hiring plan.

By Mihailo Bozic
Candidate Data: What It Includes and How Recruiters Use It
Recruiter Playbooks

Candidate Data: What It Includes and How Recruiters Use It

Learn what candidate data includes, six practical ways to use it in recruiting, and how to keep your candidate records accurate, fair, and lawful.

By Mihailo Bozic
AI Recruiting: A Practical Guide for Recruiters
Hiring TrendsRecruiter Playbooks

AI Recruiting: A Practical Guide for Recruiters

AI recruiting explained for working recruiters: what AI does at each hiring stage, the risks, the U.S. rules to know, and how to start safely.

By Mihailo Bozic
Campus Recruiting: How to Plan a Season That Ends in Hires
Recruiter Playbooks

Campus Recruiting: How to Plan a Season That Ends in Hires

Plan campus recruiting from school choice to start date: pick schools, set offer deadlines schools accept, track intern conversion, and know agency rules.

By Mihailo Bozic
B2B Technographic Data: A Practical Guide for Recruiters
Recruiter Playbooks

B2B Technographic Data: A Practical Guide for Recruiters

Learn what B2B technographic data shows, how providers collect it, and how recruiters can use it for business development, sourcing, and vetting providers.

By Mihailo Bozic
Rotto

Built for recruiters who work from a brief, not a boolean.

rotto.com