Recruitment Testing: What Scores Can—and Cannot—Tell You

Recruitment testing uses structured tasks or questionnaires to assess applicants before a hiring decision. A test might check spreadsheet skills, numerical reasoning, responses to workplace situations or personality traits. It may be scored by a person, fixed rules or AI.

The useful question is not simply whether someone passed. It is what the score measures, how that relates to the job and what decision the employer makes with it. A neatly presented score is not, by itself, evidence that someone will perform well at work.

Match the test to the question

The US Equal Employment Opportunity Commission (EEOC) distinguishes cognitive tests, sample job tasks, personality and integrity tests, physical ability tests and other selection procedures. For everyday recruitment, these distinctions help explain what an assessment is trying to establish:

Test type	What it examines	What to check before relying on it
Skills or job-knowledge test	Knowledge or ability in a particular function, such as arithmetic or software use	Does the content match what this role requires on entry?
Work sample or simulation	Performance on a task resembling the work	Are the task, tools and difficulty representative of the job?
Cognitive or aptitude test	Reasoning, memory, numerical ability or similar capabilities	What evidence connects the measured ability to this position?
Personality or integrity questionnaire	Traits, dispositions or likely conduct	What does the scale actually measure, and why is it relevant to this role?
Physical ability test	Ability to perform physical tasks, strength or stamina	Does the demand reflect the work rather than an unnecessarily harder challenge?

A work sample is not automatically the right choice. The US Office of Personnel Management (OPM) says these tests are appropriate when applicants are expected to possess the assessed competencies on entry. If the employer will teach the activity after hiring, testing mastery beforehand may be inappropriate.

For example, asking an experienced analyst to investigate errors in a spreadsheet can make sense. Requiring an entry-level trainee to master an unfamiliar internal system may test prior exposure rather than the capability the employer needs.

For the broader workflow of choosing an assessment and defining its rubric, see what to measure and how to score candidate assessments.

Read the score before acting on it

Before using a result to reject, rank or shortlist someone, establish three things:

1. The scale: Is this a count of correct answers, a rubric score, a comparison with a reference group or a model-generated prediction?
2. The decision rule: Does the result trigger rejection, determine ranking or provide evidence for a later interview?
3. The justification: Why is this threshold or ranking appropriate for this job?

Consider a fictional spreadsheet exercise scored out of 100. An applicant receives 80. That does not mean they can perform 80% of the job. The total might combine formula accuracy, explanation quality and speed; different weightings could produce a different result.

A useful scoring record would show those components and explain which errors matter. If checking accuracy is essential but speed is not, a speed-heavy total needs scrutiny.

For employers, validation means evidence supporting the test’s use for the position and purpose—not simply that the supplier tested its software. The EEOC’s employer guidance recommends proper validation and makes clear that vendor documentation does not remove the employer’s responsibility.

For candidates, ask how the result will be used. You may not receive proprietary scoring details, but knowing whether the assessment is a knockout stage or one input changes how you understand the process.

Separate AI scoring from evidence of quality

“Online” does not mean “AI.” A multiple-choice test can use a fixed answer key; a written response can be scored by a recruiter. AI may instead evaluate answers or generate a suitability prediction.

Ask the employer or supplier to identify:

• The inputs used: answers, task outcomes, recordings or other data.
• The output produced: competency scores, rankings or recommendations.
• The evidence supporting that output for the intended role.
• The review route for technical failures or questionable scores.

Accessibility also requires a separate check. The US Department of Justice’s AI hiring guidance warns that games, personality assessments and facial or voice analysis can screen out qualified people with disabilities. A test should measure the relevant skill, not an unrelated sensory, manual or speaking impairment.

For HR teams, the AI hiring vendor validation checklist provides a deeper review. A polished report or a general bias claim is not a substitute for examining the assessment’s actual use.

What candidates and HR should arrange before testing

Candidates should confirm the expected duration, deadline, permitted tools—including generative AI—technical support and accommodation contact. Use authorised practice materials to understand the format, and report access problems promptly rather than silently accepting an unusable test. These questions to ask before a hiring assessment provide a practical starting point.

HR teams should provide those instructions, define scoring before reviewing responses and record technical failures separately from poor performance. After launch, review score patterns, advancement rates, complaints and withdrawals. Keep the assessment aligned with changing job requirements, as the EEOC’s testing guidance recommends.

US legal context, checked 29 October 2026: Federal anti-discrimination rules apply to covered employers’ employment testing, not only AI. Neutral tests can raise disparate-impact concerns, and the ADA requires reasonable accommodations for the known limitations of otherwise qualified applicants with disabilities unless they would impose undue hardship. The EEOC guidance explains these standards. State and local requirements may add obligations; this is information, not legal or employment advice. Confirm applicable requirements with qualified counsel.