Pre-Employment Assessment Tests: 7 Types, What Predicts Performance, and the Legal Rules

By Sparsh Goyal, Founder at The Cognitive. Published 2026-10-04 · Last updated 2026-10-05

• A pre-employment assessment is a test or structured exercise taken before an offer, such as a skills test, cognitive test, personality questionnaire, situational judgement test, work sample, job simulation or structured interview.
• In the 2022 Sackett re-analysis, structured interviews (.42), job knowledge tests (.40) and work samples (.33) predicted performance best, and cognitive tests fell from .51 to .31.
• Under the Uniform Guidelines, a group pass rate below four-fifths of the highest group's rate is generally evidence of adverse impact, and the employer, not the vendor, must then show the test is job-related.

A pre-employment assessment, also called a recruitment assessment, is a test or structured exercise a candidate completes before an offer to show they can do the job: a skills test, cognitive ability test, personality questionnaire, situational judgement test, work sample, job simulation or structured interview. The types that predict performance best are structured interviews, job knowledge tests and work samples. In the 2022 re-analysis by Sackett, Zhang, Berry and Lievens, they scored .42, .40 and .33, ahead of cognitive ability tests at .31.

That last number used to be .51. For 24 years, the standard advice was that a cognitive test was the single best predictor you could buy. The 2022 paper found that earlier estimates had over-corrected for range restriction, and it cut most validities by .10 to .20. I run The Cognitive, which runs one of these methods, the structured interview, as a live AI video interview. Read the interview section with that in mind. Everything else here comes from the research and the regulators, read on 4 October 2026.

What is a pre-employment assessment?

A pre-employment assessment is any standardised way of measuring a job-related skill, ability or trait before you decide to hire. People also call it a pre-hire assessment, an employment assessment test or simply pre-employment testing. The point is to replace a guess from a resume with evidence from the candidate's own work or answers.

The US Equal Employment Opportunity Commission (EEOC) uses a wider net. Its list of employment tests and selection procedures covers cognitive tests, physical ability tests, sample job tasks, medical inquiries and physical examinations, personality and integrity tests, criminal background and credit checks, performance appraisals and English proficiency tests. All of them fall under the same anti-discrimination rules.

Recruiters usually mean something narrower. A pre-employment assessment measures whether the person can do the work. A background check verifies who they are and what they have done. The two sit at different points in the process, and the legal rules for each differ, so keep them apart in your planning.

Most teams place the assessment after a resume screen and before the first long human interview. That is where it saves the most time: it narrows a pool of plausible resumes to the people who have shown the skill.

What are the types of pre-employment tests?

There are 7 common types of pre-employment tests. Each one measures a different thing, and each one fails in a different way.

1. Skills and job knowledge tests

A skills test checks what the candidate knows how to do: write SQL, reconcile a ledger, read a contract, type at speed. Job knowledge tests scored .40 in the 2022 research, the second-highest figure. Their weakness is narrowness. A test of Excel formulas tells you nothing about whether the analyst can explain the result to a finance director.

2. Cognitive ability tests

The US Office of Personnel Management (OPM) says cognitive ability tests "assess abilities involved in thinking (e.g., reasoning, perception, memory, verbal and mathematical ability, and problem solving)." They are cheap, quick and easy to score at volume, and they scored .31. OPM also warns that they "typically produce racial and ethnic differences larger than other valid predictors of job performance". That makes them the test type most likely to fail the 4/5ths rule covered below.

3. Personality questionnaires

Personality tests measure tendencies such as conscientiousness, extraversion and emotional stability, the Big Five. OPM says they "tend to be less valid than other types of predictors such as cognitive ability tests, assessment centers, and work samples," and that their items can make it "easy for applicants to fake or distort test scores in their favor." The upside is fairness: OPM finds "few, if any, average score differences" between men and women or between racial groups. Our guide to personality tests for hiring compares the main tests and what each one predicts.

4. Situational judgement tests

OPM describes situational judgement tests (SJTs) as tests that "present applicants with a description of a work problem or critical situation related to the job they are applying for and ask them to identify how they would handle it." They scored .26. Candidates tend to like them, and OPM notes applicants "often perceive SJTs as being very fair." The limit is that an SJT records what someone says they would do, which is not always what they do.

5. Work samples

A work sample test asks applicants to "perform tasks or work activities that mirror the tasks employees perform on the job," in OPM's words. A support candidate answers a real-looking ticket. An engineer fixes a failing function. Work samples scored .33 and feel fair to candidates because the link to the job is obvious. OPM's warning is cost: they "may be costly to develop" and need someone to observe or rate the work.

6. Job simulations and assessment centers

A job simulation strings several work-like exercises together, often online: an inbox to prioritise, a customer call to handle, a short presentation. The classic in-person version is the assessment center, which scored .29 in the 2022 research. Online simulations are a form of work sample; the paper gives no separate figure for them. They take the most design effort of any type here.

7. Structured interviews

OPM calls the structured interview "an assessment method designed to measure job-related competencies," in which "all responses are evaluated using the same rating scale and standards for acceptable answers." It scored .42, the highest of any procedure in the 2022 paper. An unstructured interview, where each interviewer asks what comes to mind, scored .19. The rubric is what separates the two, as our explainer on what a structured interview is shows.

Which pre-employment tests predict job performance best?

Structured interviews, job knowledge tests and work samples predict job performance best. Here is the full picture from Table 3 of Sackett et al. (2022), next to the older 1998 estimates from Schmidt and Hunter that most hiring advice still repeats. Validity is the correlation with later job performance, where 1.0 would be perfect prediction and 0 would be none. The last column is the average score gap between Black and White candidates (Cohen's d) that the paper reports alongside each method; a bigger number means a bigger adverse impact risk.

3 things in that table change how I would build a process.

First, averages hide spread. Structured interviews had the highest mean but also a wide spread around it: the paper's lower 80% credibility value for them is .18. A badly built structured interview can predict no better than an unstructured one. The rubric, the anchors and the interviewer training decide which end you land on.

Second, no single test explains most of performance. A validity of .42 is the best on the list, and it still leaves most of the variation unexplained. That is the case for combining 2 methods with different weaknesses, which OPM also recommends: it says using interviews or biodata "in combination with cognitive ability tests is recommended to lower any potential adverse impact."

Third, the old top 3 lost the most. Work samples fell .21 and cognitive tests fell .20, while structured interviews fell only .09 and moved to first place. If your process leans on one timed aptitude test because it was "the best predictor", that advice rests on the 1998 figures, and the same test carries the largest group gap on the list.

Is pre-employment testing legal? Adverse impact and the 4/5ths rule

Pre-employment testing is legal in the US when every test is job-related and consistent with business necessity. The risk is adverse impact: a test that passes one group at a much lower rate than another. This is a summary for planning, not legal advice. Check your process with employment counsel in each place you hire.

The four-fifths rule

The federal Uniform Guidelines on Employee Selection Procedures (UGESP), at 29 CFR 1607.4(D), set the usual test. A selection rate for any race, sex or ethnic group "which is less than four-fifths (4/5) (or eighty percent) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact."

Here is the arithmetic. Say 120 men and 80 women take a test. 72 men pass (60%) and 24 women pass (30%). Divide the lower rate by the higher: 30 ÷ 60 = 0.50. That is well under 0.80, so the test shows adverse impact. For the ratio to reach 0.80, at least 39 of the 80 women would need to pass (48.75% ÷ 60% = 0.81).

To run the check on your own process:

1. Record, for each test step, how many people from each group took it and how many passed. Collect demographic data voluntarily and keep it away from the people making decisions.
2. Work out each group's pass rate.
3. Divide every group's rate by the highest group's rate.
4. Flag any ratio under 0.80, and look harder at small groups, where a few people can swing the number.

The rule is a rule of thumb, not a safe harbour. The same section says "smaller differences in selection rate may nevertheless constitute adverse impact, where they are significant in both statistical and practical terms."

What happens when a test shows adverse impact

Adverse impact does not make a test illegal on its own. The EEOC says the employer must then show the procedure is job-related and consistent with business necessity, and it can still lose if a less discriminatory alternative was available. UGESP section 1607.5(A) names 3 ways to show validity: "criterion-related validity studies, content validity studies or construct validity studies."

The vendor's paperwork does not move that burden. The EEOC says validity documentation from a test publisher may be helpful, but the employer "is still responsible for ensuring that its tests are valid under UGESP." The EEOC's own example shows the stakes: after Dial Corp. brought in a strength test for entry-level production jobs, women's share of hires fell from 46% to 15%, and the test was found to violate Title VII.

Disability, age and AI tools

• Disability (ADA). The EEOC says an employer "may not ask questions about disability or require medical examinations until after it makes a conditional job offer." You also owe reasonable accommodation in how a test is given, such as extra time or a reader.
• Age (ADEA). The Age Discrimination in Employment Act covers people aged 40 and over, so check pass rates by age band too.
• Automated tools in New York City. Local Law 144 bars employers from using an automated employment decision tool unless it had a bias audit within one year of use, the audit information is public, and candidates got notice.

Rules on AI hiring tools are changing in other states and countries. Check the ones that apply where you hire before you switch on any automated scoring. Our roundup of AI screening software compliance laws lists the main ones.

How do candidates experience pre-employment tests?

Candidates judge a test by whether it looks like the job. OPM's guidance reflects that: SJTs and work samples are seen as fair because the link to the work is plain, while personality items can look unrelated to the role. A test that feels arbitrary costs you the strongest candidates first, because they have other options. Our guide to candidate experience covers the wider process.

4 habits keep completion up without weakening the signal:

• Say what it measures and how long it takes in the invitation, before the candidate clicks.
• Use 1 or 2 tests, not a battery. Every extra test is another place a good candidate can quit.
• Offer accommodations up front, not after someone has struggled. The ADA requires it anyway.
• Close the loop. Tell people the outcome, and give a reason where you can.

Vendor completion rates are self-reported and measured in different ways, so I would not compare them. Run your own pilot and count how many invited candidates finish.

How do you choose a pre-employment assessment?

Choose the assessment from the job, then check it against validity, adverse impact and candidate time. These are the 5 questions I would answer first.

What does the job actually need on day 30?

Write down the 3 or 4 things a new hire must do well in their first month, then pick the test type that observes them most directly. If the job is writing code, a work sample beats a personality test. If the job is calming angry customers, a role-play or structured interview beats a typing test.

Does the test type have evidence behind it?

Start from the 2022 figures. Structured interviews (.42), job knowledge tests (.40) and work samples (.33) have the strongest record, while general personality (.19) and unstructured interviews (.19) have the weakest. A vendor's own validity claim is worth reading, but ask what jobs it was measured on.

What will the pass rates look like by group?

Cognitive tests carry the largest group gap in the 2022 table (d = .79), and integrity tests one of the smallest (.10). Run the 4/5ths check on a pilot group before rollout, and again every few months once live.

How much candidate time does it take?

Add up every test and interview in the process, not just the new one. If the total is longer than the job deserves, cut the weakest predictor first.

Who reads the result, and what do they do with it?

Name the person who reviews each result and the score or evidence that moves someone to the next round. A test nobody acts on is just a delay. Our free interview rubric generator and scorecard generator help you write those anchors down before the first candidate starts.

Where does a live AI interview fit next to a test?

A timed test checks what a candidate knows. A structured interview checks how they reason when someone asks a follow-up, which is why it tops the validity table. The catch has always been interviewer time, because every structured interview needs a trained person to run and score it.

The Cognitive runs that first structured interview as a live, two-way AI video interview of 10 or 20 minutes, in 9 languages. Candidates book their own slot without creating an account, and the invitation tells them the interview is run by AI. The rubric is fixed per role, and the questions adapt live to each answer and to the candidate's resume. The AI interviewer page shows a full report.

The recording above is a 5-minute live AI interview for a go-to-market role. The AI asks the candidate about an outbound email campaign, then follows up on reply rates and on how the leads were qualified.

Afterwards the report gives a 1 to 5 score per rubric criterion, overall written feedback, a weighted score out of 100 and a suggested verdict, with the transcript and recording behind them. The AI also probes up to 5 claims from the resume and marks each one verified, refuted or unclear, with evidence from the interview. Nothing is rejected automatically; a person decides.

Here is what it does not replace. It has no code editor, so keep a coding platform for hands-on engineering tests. It is not a cognitive or personality test. And we have not published a validity study or a bias audit, so validate it on your own roles the way you would any structured interview: compare scores with later performance and run the 4/5ths check.

Run your first-round structured interview with AI Set the rubric once, then hear every candidate answer it in a live AI video interview. Start free

Which pre-employment assessment tools sell each test type?

Most vendors specialise. This table maps the 7 types to tools whose own sites, read on 3 October 2026, list that type. Our full comparison of talent assessment tools covers pricing models and AI interview features for each, and our TestGorilla alternatives guide compares the big test libraries.

Who is The Cognitive built for?

The Cognitive is built for teams whose bottleneck is the first structured interview, and who often also need to find the candidates in the first place. Its AI sourcing tool searches ~900M public profiles from a plain-English brief, and the people you import can be invited to the live AI interview in the same role. AI Sourcing starts at $49/month and AI Interview at $99/month.

It works alongside the tests you already trust:

• Keep your skills or coding test for hands-on proof, and use the interview for how a candidate explains their choices.
• Keep a short aptitude test if your validation supports it, and pair it with the interview, as OPM suggests, to lower adverse impact risk.
• Keep your ATS. The Cognitive has 60+ ATS integrations and writes a note with the score, summary and report link back to candidates imported from it.

There is a free trial with 100 sourcing credits.

Put one open role through a structured AI interview Write the rubric, invite candidates, and read scored reports with the recording behind every score. Start free

Sources

• Sackett, P. R., Zhang, C., Berry, C. M., and Lievens, F., Revisiting meta-analytic estimates of validity in personnel selection: Addressing systematic overcorrection for restriction of range, Journal of Applied Psychology 107(11), 2040 to 2068, 2022. Table 3 (validity, SD, credibility value, Black-White d), read 4 October 2026.
• O'Shea, P. G., and Luscombe, A. F., Is Cognitive Ability the Best Predictor of Job Performance?, The Industrial-Organizational Psychologist (SIOP), read 4 October 2026.
• Uniform Guidelines on Employee Selection Procedures, 29 CFR 1607.4(D) and 29 CFR 1607.5(A), read 4 October 2026.
• US Equal Employment Opportunity Commission, Employment Tests and Selection Procedures, read 4 October 2026.
• US Office of Personnel Management, Cognitive Ability Tests, Personality Tests, Situational Judgment Tests, Work Samples and Simulations and Structured Interviews, read 4 October 2026.
• NYC Department of Consumer and Worker Protection, Automated Employment Decision Tools (Local Law 144), read 3 October 2026.
• Vendor test catalogs (TestGorilla, HackerRank, CoderPad, Codility, Criteria, Wonderlic, SHL, Bryq, Vervoe, Harver), read 3 October 2026; links in our talent assessment tools comparison.

All posts · thecognitive.io