The honest answer
The vast majority of free online IQ tests are not psychometrically valid measures of intelligence. They are entertainment products designed to flatter users, maximise social sharing, and in many cases, gate results behind a paywall after generating enough emotional investment.
This is not a minor calibration issue. Tests that report average scores of 120–130 to general populations are producing results that are structurally dishonest. They are not measuring what they claim to measure.
Why most online tests inflate scores
The incentive structure is clear: people who score well share their results; people who score poorly do not. A test that tells most users they are "significantly above average" will be shared far more widely than an accurate test. This creates evolutionary pressure toward inflation among tests competing for web traffic.
Many online tests also use a small number of simple questions weighted toward visual-spatial puzzles — the type of question where the average person performs relatively well. Omitting verbal, numerical, and working memory components makes the test easier and produces higher scores.
What makes an IQ test valid
A valid IQ test must: (1) be normed on a large, representative population sample, (2) include multiple cognitive domains (verbal, quantitative, spatial, working memory), (3) use Item Response Theory or equivalent calibration methodology, (4) report confidence intervals alongside point estimates, and (5) have published reliability and validity data.
The gold standard clinical IQ tests — WAIS-IV, Stanford-Binet 5, WISC-V — satisfy all of these criteria. They are administered by licensed psychologists, take 60–90 minutes, and produce scores with documented reliability coefficients above 0.95.
Red flags to watch for
Red flags that indicate an untrustworthy online IQ test include: reporting results without confidence intervals, not disclosing the norming sample, reporting that most users score "above average," using fewer than 20 questions, completing in under 5 minutes, gating score access behind payment, and reporting scores above 130 with high frequency.
Signs of a more honest test
Better online tests: report confidence intervals with every score, explain their norming methodology, include diverse question types across cognitive domains, take at least 20–30 minutes, use adaptive question selection (harder questions for better performers), report an average score close to 100 for general populations, and explicitly note that results are estimates, not clinical assessments.
Online tests vs clinical assessment
Clinical IQ assessment by a licensed psychologist differs from any online test in several important ways: it is individually administered (allowing the examiner to note testing conditions, motivation, and anxiety), it includes subtests measuring specific cognitive components, the examiner can adapt the administration to the individual, and the norms are regularly updated to population shifts. None of these advantages are available in an automated online format.
The practical difference in accuracy is substantial. Clinical assessment typically has reliability coefficients of 0.95–0.97, meaning measurement error is roughly 3–5 IQ points. Well-designed online tests may achieve 0.85–0.90, implying measurement error of 5–8 points. Poor online tests may have reliabilities below 0.70, implying errors of 10+ points — making the scores essentially uninformative.
What you can trust
For genuine clinical decisions (educational placement, disability assessment, Mensa applications), only proctored, clinically validated tests administered by qualified professionals are acceptable. For general orientation — getting a rough sense of your cognitive strengths and weaknesses — a well-designed online test with honest norming can provide useful signal, interpreted with appropriate uncertainty. Treat any online score as a range estimate (±10 points) rather than a precise measurement.