What accuracy means for an IQ test
Psychometrics splits the question into two parts. Reliability asks whether the test measures consistently: would the same person get a similar score on a second sitting, and do items within the test agree with each other? Validity asks whether it measures the thing it claims to: does performance relate to other established measures of reasoning?
Both are reported as numbers, which makes them checkable. Internal consistency and test-retest correlations above roughly .85 are considered strong for a group-administered reasoning test. A test that publishes neither figure has not made an accuracy claim you can evaluate.
The numbers behind the Testora IQ test
Testora IQ items are calibrated on a normative sample of 18,400 completed sittings. Internal consistency is .91, test-retest correlation across sittings separated by at least 30 days is .86, and the standard error of measurement is 4.5 points.
Those figures set a realistic expectation: the score is a good estimate of your standing, reported with a confidence interval of about plus or minus seven points at 90% confidence. It is not a measurement to the point, and no online instrument is.
Where online testing genuinely loses ground
An individually administered clinical battery has advantages an online test cannot replicate, and pretending otherwise would be dishonest.
- Supervision — an examiner confirms identity, controls the environment and notices fatigue, distraction or misunderstanding of instructions.
- Behavioural observation — how someone approaches a problem is often as informative as whether they solve it, and only a live examiner sees that.
- Breadth — clinical batteries sample verbal comprehension and processing speed with subtests that need spoken interaction or precise timing hardware.
- Standardised conditions — screen size, input device, notifications and interruptions vary from home to home.
Where a good online test holds up well
For non-verbal reasoning measured under time pressure, a calibrated online test performs close to its supervised equivalent. Item difficulty can be estimated from far larger samples than most clinical instruments ever collect, and scoring is applied identically to everyone with no examiner drift.
For the purposes most people actually have — understanding relative reasoning strengths, choosing study strategies, satisfying curiosity, screening before a formal assessment — that is sufficient.
How to spot an unreliable online IQ test
Most of the tests that give the category a bad name share the same handful of tells.
- No reported reliability, sample size or standard error anywhere on the site.
- A precise score with no confidence interval and no percentile.
- Scores that come out flatteringly high for almost everyone, which usually means no real norming.
- Payment required before you can see any score at all.
- Claims of clinical or diagnostic value, or of measuring potential and future success.
How to get the most accurate result you can at home
Conditions account for several points of variance, and you control most of them. Sit the test when rested rather than late at night, use a laptop or desktop rather than a phone, silence notifications, work in one uninterrupted block, and do not look answers up — a score inflated by searching tells you nothing you can use.
If you want a stability check, retake after at least 30 days rather than immediately, and expect the two results to land within the confidence interval rather than to differ meaningfully.
Frequently asked questions
- Are online IQ tests accurate?
- A properly normed online IQ test gives a reliable estimate of non-verbal reasoning with a confidence interval of a few points. It cannot replace a supervised clinical assessment.
- Can an online IQ test diagnose anything?
- No. Diagnosis of learning differences or cognitive impairment requires an individually administered battery interpreted by a qualified psychologist.
- Why did my score change on a second attempt?
- Differences of a few points are measurement error plus condition effects such as sleep, stress and familiarity with item formats, not a change in ability.