Are online IQ tests accurate?
Some are genuinely informative. Most are decorative. The gap between them is not a matter of opinion - it comes down to three properties you can check.
On this page
Three questions that settle it
"Accurate" is too loose a word to argue about. Psychometrics splits it into three properties, and a test can be strong on one and useless on another.4
1. Reliability - would it give you the same answer twice?
If the same person takes the test twice under the same conditions, how close are the two scores? A test with poor reliability is measuring noise, and nothing else about it can rescue that. Reliability is reported as a coefficient between 0 and 1; the WAIS-IV reports about .98 for Full Scale IQ.3 A short online test in the .80s is doing well.
2. Validity - is it measuring the thing it claims to?
A reliable test can reliably measure the wrong thing. Validity is established by showing the scores behave as a measure of the construct should: correlating with established instruments, predicting outcomes the construct ought to predict, and having the expected internal structure.6 This is where most free online tests have nothing to offer, because nobody has ever checked.
3. Norming - is the scale anchored to anything?
An IQ number is a rank within a reference population. That requires a standardisation sample. A test that has never been administered to a defined sample cannot know whether your raw score of 24 out of 30 is the 60th or the 95th percentile - it can only assume. Most online tests, including this one, are weakest here.
What the research actually shows
The useful evidence comes from the International Cognitive Ability Resource, a public research project built specifically to test whether an open, brief, online battery can work. The initial validation drew on 96,958 participants from 199 countries; the 16-item short form reached an internal consistency of α = 0.81.1
The direct comparison came later. Young and Keith administered the ICAR16 alongside the WAIS-IV to 97 university students. The correlation between ICAR16 scores and WAIS-IV Full Scale IQ was r = .81, and between the latent general factors, .94.2
Read that result carefully
An r of .81 is high, and it comes with real caveats. The sample was 97 university students - small, and restricted in range, which usually deflates a correlation rather than inflating it. But a student convenience sample also cannot be assumed to represent the general population, and the study's own authors called for replication in larger samples.
What it does establish: a short, unsupervised, browser-based reasoning test is capable of capturing much of what an hour-long supervised battery captures. What it does not establish: that any particular website has built one.
The critical distinction is between the format and the implementation. Being online is not the problem. Being unvalidated is.
Online versus supervised: a comparison
| Supervised battery (e.g. WAIS-IV) | A well-built online test | A typical free online test | |
|---|---|---|---|
| Reliability | ~.98 | ~.80-.90 | Usually unknown |
| Standard error | ~2 points | ~4-6 points | Unknown |
| Norms | Large stratified sample | Sometimes; often provisional | Usually none |
| Abilities covered | Broad, multiple indices | Narrow, reasoning-focused | Varies |
| Conditions | Invigilated, standardised | Unsupervised | Unsupervised |
| Time | 60-90 min | 15-35 min | 2-15 min |
| Cost | Several hundred, via a psychologist | Free or low | Free, or paywalled at the result |
| Clinical standing | Yes | None | None |
The honest summary: a good online test tells you roughly which part of the distribution you are in. It will not distinguish 112 from 118, and it cannot support a diagnosis, an accommodation request or an application to anything.
Eight-point checklist
Apply this to any online IQ test, including this one.
- Does it publish its method? If there is no page explaining how items were built and how the score is computed, there is nothing to evaluate.
- Does it report a margin of error? Every measurement has one. A bare number with no interval is overstating its precision - always.
- Does it say where its norms come from? And if they are provisional, does it say so?
- Is it long enough? Below about 20 items, a score cannot be precise regardless of anything else. Five-question tests are entertainment.
- Does it cite anything? Real references to real papers, not "based on scientific research" with nothing attached.
- Does it state its limitations? A test that lists no weaknesses has not looked for any.
- Does it ask for money to see your score? The single most reliable red flag.
- Does the score seem too flattering? Tests that hand out 130s to most takers are optimising for sharing, not measurement.
Red flags
- A paywall at the result. You have done the work; the score is then held hostage. This model has no incentive to be accurate, only to seem tantalising.
- "Certified" or "official" claims. There is no certifying body for online IQ tests. The word is decorative.
- Impossibly high scores. A test reporting 145+ to a meaningful share of takers is not measuring the tail; it is miscalibrated.
- No mention of error, ever.
- Instant results from very few questions. Precision requires items. There is no way around that.
- Email required before the score. The product is your address.
How this site scores against its own checklist
It would be poor form to publish that list without applying it here.
| Criterion | This test |
|---|---|
| Method published | Yes — in full, including the scoring model |
| Margin of error reported | Yes — a 95% interval on every result |
| Norms disclosed | Yes, and they are the weak point: provisional and rational, not empirical |
| Length | 30 items, about 30 minutes |
| References | 17 sources on the methodology page |
| Limitations stated | Yes — nine of them |
| Paywall | None. No account, no email |
| Score inflation | Scores are capped at 135 because a 30-item test cannot resolve beyond it |
The row that matters most is the third one. This test has no standardisation sample of its own, so the absolute number it gives you is the least trustworthy part of the report - which is exactly why it comes with an interval attached and a page explaining why.
Common questions
Are free online IQ tests accurate?
A minority are reasonably accurate; most are not. The research shows the format can work - a short online battery correlated r = .81 with the WAIS-IV in one study - but that finding applies to a specific validated instrument, not to online tests in general. Judge each one on whether it publishes a method, reports error, and discloses its norms.
How much can an online IQ score differ from a real one?
Expect a band rather than a point. A good short online test has a standard error of roughly 4 to 6 IQ points, so a 95% interval of about ±9 to ±12. On top of that, unsupervised conditions and provisional norms can add systematic bias in either direction.
Which online IQ test is the most accurate?
The most defensible ones are those derived from published, validated item banks - the ICAR work being the clearest example - and those that publish their scoring model and error. Be suspicious of any ranking, including this sentence: nobody has run a controlled comparison of the popular free tests against a supervised battery.
Can an online IQ test diagnose anything?
No. Diagnosis requires a qualified clinician, a standardised individually administered instrument, and evidence beyond a single score. No online test can do this, and any that implies otherwise should be disregarded entirely.
References
- Condon, D. M., & Revelle, W. (2014). The International Cognitive Ability Resource: Development and initial validation of a public-domain measure. Intelligence, 43, 52-64. doi:10.1016/j.intell.2014.01.004
- Young, S. R., & Keith, T. Z. (2020). An examination of the convergent validity of the ICAR16 and WAIS-IV. Journal of Psychoeducational Assessment, 38(8), 1052-1059. doi:10.1177/0734282920943455
- Wechsler, D. (2008). Wechsler Adult Intelligence Scale - Fourth Edition: Technical and Interpretive Manual. Pearson.
- American Educational Research Association, American Psychological Association, & National Council on Measurement in Education (2014). Standards for Educational and Psychological Testing. AERA.
- Embretson, S. E., & Reise, S. P. (2000). Item Response Theory for Psychologists. Lawrence Erlbaum Associates.
- Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52(4), 281-302.
Related reading
How this test is built
Item construction, the scoring model, precision and limitations.
IQ score ranges and percentiles
What each score corresponds to, and how rare it is.
Practice questions
Worked examples of each question type, with explanations.
What is an average IQ?
Why 100 is the average by construction rather than by measurement.
Find out where you land
Thirty questions, about 30 minutes, and a score with the uncertainty attached.
Start the test30 questions · about 30 minutes · no sign-up