How does this IQ test work?
This IQ test is 24 reasoning puzzles, scored on our own scale, with no words to decode beyond a one-line instruction. Each question is built in your browser the moment you press Start, from a random seed, so no two people are likely to see the same set and nobody can memorize the answers from a forum. The seed is shown as a test code on your result; the same code always rebuilds the same 24 puzzles.
Questions come in six rounds of four, from easiest to hardest. You can go back, change an answer or skip. Time is reported on the result and has no effect on the score. Automated tests solve every puzzle a second, independent way: for 3,000 seeds, every question has six different options and exactly one that fits the rule.
What do the four puzzle types measure?
The four types are number sequences, pattern grids, odd one out and balance logic. Pattern grids are 3×3 matrices with the bottom-right piece missing, the format made famous by Raven's Progressive Matrices; the rules come from the family Carpenter, Just and Shell identified in their 1990 analysis of that test. In the last balance round, three unknown weights hide in three equations, and the quickest route is to add the scales together.
| Type | Early rounds | Last rounds | Leans on |
|---|---|---|---|
| Number sequences | Add the same amount each time | Doubling, squares, two interleaved series, alternating operations | Proposing and testing a rule |
| Pattern grids | Simple row rules, such as a feature that stays constant | Three rules stacked at once | Working memory |
| Odd one out | The shape differs | Even or odd count of shapes, then a mirror-image L among rotations | Spotting the hidden rule |
| Balance logic | Scales that balance | Three unknown weights in three equations | Algebra without letters |
Source: The Daily Summit.
How is the score calculated?
A correct answer earns as many points as its round number, so the maximum is 84, and later, harder answers count for more. Your share of the 84 is placed on a scale built to look like the familiar IQ scale: an average of 100 and a standard deviation of 15.
The two constants are our design targets: we built the difficulty curve so a typical adult earns about 52% of the points, with most people within 16 points of that. Scores are held between 70 and 145, and the range shown is your score plus or minus 5.
We have not tested the puzzles on a large, representative group of people, which is what a clinical test publisher does before printing a score table. Hence the line on every result: Scored on The Daily Summit's own scale; not a clinical IQ assessment.
| Score | Band | What it usually reflects | Expected share |
|---|---|---|---|
| 130 to 145 | Exceptional on this test | Solved most of the hardest items, including mirror images and three-rule grids. | about 2% |
| 115 to 129 | Well above our average | Comfortable with every type; lost points mainly on the last two levels. | about 14% |
| 100 to 114 | Average to above average | Solid on the first half; mixed results once two or three rules combine. | about 34% |
| 85 to 99 | Average to below average | Most early items right; the later, multi-step puzzles were harder. | about 34% |
| 70 to 84 | Room to grow on these puzzles | Often a sign of rushing, distraction or unfamiliar puzzle formats, which practice changes quickly. | about 16% |
Source: The Daily Summit. Expected share is what a bell curve with mean 100 and SD 15 would give; it is not measured.
What does the AI breakdown send?
Press Explain my result and the page sends your score, the number right and asked in each puzzle type, and your time in seconds. It does not send your answers, a name or an email; as with any web request, the server sees your IP address. A small language model on Cloudflare's Workers AI turns those numbers into a short note on your strongest and weakest puzzle types. Otherwise the test runs in your browser and nothing is saved.
The model is told not to diagnose anything or describe your score as a clinical IQ. If the service is unreachable, the page writes the note from fixed rules instead.
Is an online IQ test a clinical IQ?
No. A clinical IQ comes from a test such as the Wechsler Adult Intelligence Scale (WAIS-5, 2024), given in person by a qualified psychologist. Its score tables come from testing a large sample chosen to mirror the population.
That sample is the whole point. In 1996 a task force of the American Psychological Association, led by Ulric Neisser, described how IQ scores are converted to a scale with a mean of 100 and a standard deviation of 15, which puts about 95% of people between 70 and 130. A score of 115 only means "higher than about 84% of people" if the test was normed on those people.
Even good norms go stale. The political scientist James Flynn showed in 1987 that average scores on the same tests had risen sharply within a generation in all 14 nations he examined, which is why publishers renorm their tests every decade or two. An online quiz with no norms at all, ours included, can tell you how you did on its puzzles and how that compares with its own design, and not much more.
How do you get a fair result?
- Sit somewhere quiet for 15 minutes with notifications off. Interruptions cost more points than any puzzle.
- Use paper for sequences and scales. Writing down the gaps between numbers is what experienced solvers do.
- Check every option on grid questions before choosing. Wrong options often differ from the answer in a single feature.
- Skip and come back if a puzzle stalls you for more than a minute or two. The Back button keeps your answers.
Expect to score higher on a second attempt: a 2018 meta-analysis by Scharfen, Peters and Holling, covering 122 studies, found that scores on cognitive ability tests rise with retesting and level off after about the third attempt. Your first score is the more honest one. To track change over time, a quick memory test, reaction time test or typing test suits better.
Sources
- Neisser and others: Intelligence, knowns and unknowns, American Psychologist (1996)
- APA Dictionary of Psychology: intelligence quotient
- American Psychological Association: Intelligence
- Pearson: Wechsler Adult Intelligence Scale, Fifth Edition (WAIS-5)
- Carpenter, Just and Shell: What one intelligence test measures, Psychological Review (1990)
- Flynn: Massive IQ gains in 14 nations, Psychological Bulletin (1987)
- Scharfen, Peters and Holling: Retest effects in cognitive ability tests, a meta-analysis, Intelligence (2018)
Last reviewed November 9, 2026. For entertainment and practice. Scored on The Daily Summit's own scale; not a clinical IQ assessment. It is not a diagnosis of anything.