Candidate Screening Software: How to Choose and Set It Up

Candidate screening software sorts job applicants against your job requirements before a person reads them. It helps when applications pile up faster than your team can review them.

These tools ask knockout questions, rank resumes, or run a short chat, video, or skills screen. The bigger choice is which decisions you let the software make on its own.

What candidate screening software does

Candidate screening software, also called applicant screening software, filters and ranks the people who apply. It may be a separate product or a set of features inside your applicant tracking system (ATS).

AI already plays a part in this work for HR teams. A 2025 SHRM survey of HR professionals found that 43% use AI to support HR work, with the majority using it for recruiting. Reviewing or screening resumes was one of the named uses. The survey covered HR professionals in general, not recruiters alone.

So before you buy anything, check what your current system can already do. Your ATS may already offer knockout questions and basic ranking. Our guide to recruitment software shows where screening fits in the wider stack. A separate tool makes more sense when your volume needs something the ATS can't handle, such as text screening for hourly roles.

Say a staffing team is hiring certified nursing assistants (CNAs) for a group of nursing homes. The job requirements are an active CNA certification, night or weekend availability, and recent hands-on patient care. A search like this can draw a large pile of applicants. Much of the early sorting turns on facts a candidate can state in one answer.

Types of candidate screening tools

Candidate screening tools fall into five types, and each handles a different part of the job. The table shows what each type does, what it can miss, and where a person should step in.

Resume ranking is what people often mean by resume screening software. For the CNA search, though, knockout questions and a short text screen would do most of the early work. That's because the requirements are plain facts.

Decide what software should screen and what people should

Software should check facts that are plainly yes or no and sort everyone else. People should make any call that needs judgment, and a person should review applicants before anyone is rejected on a ranking.

Yes-or-no facts include a required certification, shift availability, and legal work eligibility in the U.S. as the candidate states it. A candidate can answer each in one word, so the software can act without guessing. Judgment calls are different. Whether a career gap matters, or whether related experience counts, needs a person who can weigh the details.

Ranking sits in between. A ranked list helps you decide who to read first, but a low position isn't proof that someone is unqualified. A 2026 SHRM article quoted an interviewer-training firm co-founder on this point. Relying on an automated score without discussion can obscure important context, and the same expert called it a compliance risk.

In the CNA example, candidate screening software can screen out people who say they have no CNA certification. It can do the same for people who can't work nights or weekends. Whether two years of home care counts as recent patient care is a judgment call, so that applicant goes to a person. The first call then covers what software can't check, and our phone screening questions can help you plan it.

How AI resume screening ranks applicants

AI resume screening compares the text of each resume with the job's requirements and scores how closely they match. Tools differ in the details, but the result is an ordered list rather than a yes or no.

The ranking is only as good as the words it works from. One CNA might write "assisted residents with daily living" and another "provided direct patient care." Both may have the same experience, yet they can rank far apart. A missed keyword says something about how a resume was written, not about the person, so a low rank is a reason to look.

Ask how the ranking works before you rely on it

Before you turn on ranking, ask the vendor or your ATS provider what the score is based on. Ask, too, whether you can see why each applicant ranked where they did. A tool that shows its reasoning lets you spot a bad match quickly. A tool that only shows a number leaves you rereading resumes anyway.

For the CNA search, the ranking should weigh recent patient care and shift availability, not job titles or years in the field. Write those requirements down first, as you would for an interview scorecard. That way you and the hiring manager or client agree on what the ranking should reward. Our guide to AI recruiting explains how these models work in more detail.

Set up knockout questions that don't reject good people

Good knockout questions ask only about true job requirements, in words a qualified person can't misread. Each one rejects people automatically, so a bad question can cost you candidates without anyone noticing.

Start with the requirements the job can't be done without. A preference, such as years in the field, belongs in the ranking or the first call instead. Then word each question the way your candidates describe themselves. Leave out anything that could screen people out for reasons the work doesn't call for.

In the CNA example, the knockout questions might look like this:

• Certification: "Do you hold an active certified nursing assistant (CNA) certification?"
• Shifts: "Can you work night shifts, weekend shifts, or both?"
• Recent experience: "Have you provided hands-on patient care in the last two years?"

Each question has one meaning and checks one requirement. The team could also move the experience question into the ranking if it wants to see career returners. A person can then judge whether older experience still counts.

Automatic rejection rules carry legal weight too. In 2023, the U.S. Equal Employment Opportunity Commission (EEOC) settled a suit alleging one employer's software rejected women 55 or older and men 60 or older. The EEOC said the employer had programmed those rules into its own application software.

A knockout question is the same kind of setting, so it deserves the same care. Before launch, work through our compliance checklist and ask HR or employment counsel about the rules where you hire.

Test your setup on past applicants

Run your candidate screening software on a batch of past applicants, including people you hired, before you use it on live ones. If it rejects someone who turned out to be a strong hire, the setup needs fixing.

Pick applicants from the last few searches for the same role. Include your best hires along with people you turned down, and run them through the questions and ranking as you plan to configure them. Then read every rejected profile from the group you'd have hired, and find the rule that caught each one.

What to monitor after launch

Once candidate screening software is live, keep checking it, because applicant pools and job requirements change. These signals show whether it's still working as you set it up:

• Rejection rate at each step: A sudden jump after a knockout question can point to confusing wording.
• A sample of rejected resumes: Have a person read a handful each week to make sure good people aren't being turned away.
• Candidate drop-off: If people start the application and quit at a video or skills screen, that step may be too long for the role.
• Complaints: A confusing question can show up in what applicants say about the process.

Reading rejected resumes takes the most time of the four, but it catches problems the numbers miss. Our guide to resume red flags can help that reviewer tell real concerns from formatting quirks.

Check who gets rejected before you trust the software

Candidate screening software saves the most time when it handles plain facts and leaves the harder calls to people. The setup choices that cause trouble are small, like one misworded question or a ranking nobody checks. A test on past applicants and a weekly look at the reject list can catch them early. So before you switch anything on, run last month's applicants through it and read every name it would have turned away.