Psychometric and skills tests are usually taken online, unsupervised, on the candidate's own device. That raises an obvious question for any employer relying on the results: can candidates cheat, and does it matter? The short answer is that some try, most ways of trying are detectable or ineffective, and a well-run process makes the results trustworthy. This article explains how cheating happens and what to do about it.
How candidates try to cheat
Getting help. A friend sits beside the candidate, or sits the test for them. This is the oldest method and still the most common.
Looking answers up. Candidates open another tab to search for an answer, or paste a question into an AI assistant. Knowledge-based questions are the most exposed; reasoning questions under time pressure much less so.
Practising the exact test. Candidates who have seen the same questions before, through a friend or a leaked copy, can score well without the underlying ability.
Guessing. Multiple-choice tests can be guessed. This is not cheating, but it does add noise to a single short test.
Does it matter?
Less than it first appears, for three reasons. First, a candidate who scores well by cheating has to sustain the performance in the interview and the job, and cannot. Second, timed reasoning tests leave little room for outside help: there is no time to search, and a helper has to be as able as the candidate would need to be. Third, the test is one input; the interview, references and a work sample catch what the test misses.
It matters most when a single unsupervised knowledge test is used as the only filter, which is also poor practice for other reasons.
How to prevent cheating
Use the platform's integrity checks. Test Candidates records how each candidate sat the test: whether they left the test tab, how long each question took and whether the pattern looks like one person working under time pressure. Every report carries these integrity checks, and you can turn on optional camera photos taken during the test so you can see who sat it.
Randomise and time-limit. Questions are drawn and shuffled so two candidates do not see the same order, and the time limit makes looking answers up impractical for reasoning tests.
Combine test types. Pair a knowledge test, which is easiest to look up, with a reasoning test or a situational judgement test, which are not. A candidate who scores far higher on knowledge than on reasoning is worth a question.
Follow up in the interview. Ask a shortlisted candidate to talk through one question from their test, or set a short live task. Someone who did the work can explain it; someone who did not cannot.
Tell candidates what you are doing. A clear note that the test is monitored and that results will be discussed at interview deters most attempts before they start.
What to do if you suspect it
Do not reject on suspicion alone. Look at the integrity flags on the report, compare the result with the rest of the assessment and raise it in the interview with a related question. A genuine candidate will have no difficulty; a cheated score falls apart quickly.
Used this way, online testing is a reliable first stage. It gives every applicant the same questions under the same conditions, shows you who to interview and leaves the final decision where it belongs: with you, after you have met them. You can see the integrity checks on a sample report and try the tests with your own candidates for free.
