Are Online IQ Tests Accurate? An Honest Answer

The verdict: a screener, not a score

An IQ score means something only relative to a reference group, measured under controlled conditions, with a known margin of error. Clinical tests meet all three conditions. Most free online tests meet none of them fully. That does not make them worthless: a well-built short test still ranks people roughly, and an extreme result says something. It does mean the number on the screen is not an IQ in the clinical sense.

The formats show the gap. Pearson describes the WAIS-5 as an individually administered clinical instrument for ages 16 to 90. Its full-scale IQ takes seven subtests and about 45 minutes, and the ten primary index subtests about an hour. A typical free web test is a few dozen puzzles taken alone at home, with no one watching the clock or the browser tabs.

What makes any IQ test credible

The field's rulebook is the Standards for Educational and Psychological Testing, published jointly since 1966 by the American Educational Research Association, the American Psychological Association and the National Council on Measurement in Education. The current edition dates from 2014 and is open access. Its core demands reduce to six questions any test should answer:

• Norms: Who is the comparison group, how many people, and when were they tested?
• Reliability: Does the test give consistent results, and what is its standard error of measurement?
• Validity: Do its scores agree with established tests such as the Wechsler or the Stanford-Binet?
• Administration: Was the test timed and supervised, with identity checked?
• Ceiling: Does it contain enough hard items to separate high scorers?
• Currency: Are the norms recent enough to avoid drift?

A test that publishes clear answers to all six has earned attention. One that publishes none has not.

The norm problem: compared with whom?

A score of 130 means better than about 98 percent of a defined population. Clinical tests define that population with care. The WAIS-IV was standardized on 2,200 people in the United States aged 16 to 90, plus an extension sample of 688 Canadians. The WAIS-5 collected new norms in 2023 and 2024 from a sample built to reflect the US population.

Most online tests compare each taker with whoever else took the test. That group is self-selected: people curious about their IQ, some of them several times over. A percentile against that crowd is real information about the crowd. It is not a percentile of the general population.

Norms also go stale. Average raw scores rose through much of the 20th century, about 3 points per decade in the United States on Wechsler tests, while Denmark and Norway later recorded reversals. A test normed long ago, or never normed at all, cannot place anyone accurately.

Reliability and ceiling: why short tests blur the top

Every test score contains error. On modern professional batteries, the reported standard error of measurement can be as low as about three points, and the confidence interval spans about 10. Short tests average over fewer items, so their error is larger, and few of them publish it.

The top of the scale is the hardest part to measure. High IQ scores are less reliable than scores near the median, and reports of scores much higher than 160 are widely treated as dubious; the WAIS-IV itself stops at 160. A short online test with few genuinely hard items runs out of ceiling long before that. Results such as 160 or 187 from a quick web quiz should be read as artifacts of the scoring, not as measurements.

Practice effects: the retake trap

Online tests usually allow unlimited retakes, and retakes inflate scores. A 2018 meta-analysis by Scharfen, Peters and Holling in the journal Intelligence pooled 174 samples from 122 studies, covering 153,185 people. Scores rose by 0.33 standard deviations from the first to the second administration, about 5 IQ points, and by 0.50 by the third, about 7.5 points. After the third attempt, the gains stopped.

That is the difference between measuring ability and measuring familiarity. Pearson restricts who may buy the WAIS-5, and psychologists ask about recent testing; an open web test can do neither. The first attempt is the only one worth reading, and even that may follow practice on a similar test elsewhere.

Online testing that does count

Being online is not the problem; supervision and norms are. British Mensa runs an online supervised IQ test, about 1½ hours for £49, that qualifies for admission. Its £15 home test, by contrast, returns only an indicative score. Mensa International's site said an online adaptive test for applicants without a national Mensa was expected to go live in 2025.

Researchers have built credible open tools as well. The International Cognitive Ability Resource is a public-domain item bank designed for social-science research, with psychometric data available for its item types. It shows that a free test can be serious, provided it discloses how it was built and what it can and cannot do.

Where our free test fits

Our free IQ test is 20 questions in 8 minutes across numerical, logic and verbal reasoning, with no sign-up. It ranks each result against the real people who have taken it, not against a textbook curve. That is its strength and its limit: the comparison group is our visitors, not a census-matched sample, and 20 items cannot fix a score to within a few points.

Used honestly, that makes it a good first step. In eight minutes it shows roughly where a person stands, and it costs nothing. If the result comes back high, the next move is a supervised test, and our guide to joining Mensa covers the options. If it comes back lower than expected, remember the margin of error and the cost of a tired evening. To interpret any number, use the IQ rarity chart and the IQ classification chart.

Red flags and green flags

• Red flag: a certified or official IQ from an unsupervised test.
• Red flag: a free test whose result is locked behind a payment.
• Red flag: scores above 160 handed out routinely.
• Red flag: no information about norms, sample size or error.
• Green flag: a named norm group, with its size and date.
• Green flag: a stated margin of error or confidence range.
• Green flag: plain wording about what the score does not mean.

For how professional tests are built and scored, see how IQ tests work and how IQ is measured.

Editorial standard: IQ figures for historical figures are retrospective estimates and are labeled as such; scores for living people appear only when the person disclosed them or a reputable outlet reported them. Corrections: contact the editors.