What 3,834 Online IQ Tests Reveal: Rushing Costs 17 Points, Spatial Reasoning Is the Weakest Skill

Key findings

1. Rushing is expensive. Test-takers who finished in under 3 minutes averaged a score of 99.6. Those who took 3–5 minutes averaged 117.1 — a 17.5-point gap. Taking much longer does not help either: the 43 people who spent more than 20 minutes averaged 109.5.
2. The sweet spot is 3–8 minutes. Almost half of all test-takers (1,824 of 3,834, or 48%) finished in that window, and they scored highest. The median time was 7 minutes 58 seconds.
3. Spatial reasoning is the weakest skill. Across 3,810 tests, people answered 58.4% of spatial questions correctly, versus 63.6% for logical reasoning, 69.2% for verbal ability, 74.4% for pattern recognition and 76.9% for memory.
4. Afternoons have a small edge. Among 2,968 tests from countries with a single time zone, people testing between 1 p.m. and 4 p.m. local time averaged 116.4, against 113.3 between 5 a.m. and 8 a.m. The difference is small but consistent with research on circadian alertness.
5. Saturday is the best day, just. Saturday tests averaged 115.9; every weekday fell between 114.1 and 114.4.
6. Online test-takers are not a random sample. The average score was 114.5, not 100. One in seven (13.3%) scored 130 or above; only 1.5% scored below 85. People who seek out an IQ test online are, on average, more confident in their abilities than the general population.

1. Speed: three to eight minutes is the sweet spot

The standard IQ-EXAMS.COM test has 25 questions and a banked timer: it starts with 45 seconds, every answer adds another 45, unused time carries over, and the test ends when the bank is empty. Within that bank each person sets their own pace, which makes it possible to look at how time spent relates to the score.

The 117 people who finished in under three minutes got 41% of questions right. Pure guessing on four-option questions would give about 25%, so they were not clicking at random, but they were clearly not reading carefully either. Their average score of 99.6 says less about their intelligence than about their patience. From three minutes upwards the picture changes completely: 3–5 minutes averaged 117.1, 5–8 minutes 115.4 and 8–12 minutes 114.1.

The slow decline after five minutes is worth a second look. It is not that thinking longer makes you worse; it is more likely that people who find the questions hard take longer, and people who find them easy move fast. The 43 people who took more than 20 minutes (109.5 on average, 54% correct) probably include a mix of very careful test-takers and people who paused the test to do something else.

The practical takeaway for anyone about to take an IQ test: read every question, but do not agonise. For a 25-question test, something between three and ten minutes is where most strong scores sit.

2. Spatial reasoning is the hardest domain

Each test has five questions in each of five domains. Comparing how many questions people got right in each domain shows a clear ranking. Spatial reasoning was the lowest-scoring domain in six of the seven months of the study, and memory the highest in all seven.

Spatial questions were answered correctly 58.4% of the time. These items are described in words rather than drawn: what solid you get when you roll a pizza-slice sector so its straight edges meet, what a tetrahedron looks like from directly above (28% got that one right), or how many lines of symmetry a square has. Memory questions, at the other end, 76.9%. Logical reasoning sits second from the bottom at 63.6%.

Two things probably drive this. Spatial items are the least like anything people practise in daily life or at school, where verbal and memory skills are exercised constantly. And a spatial question with no diagram asks you to build the picture in your head from words alone, before you can even start reasoning about it. The question bank was refreshed at the end of August 2026; after that, spatial accuracy rose from the mid-50s to the mid-60s, but it stayed at or near the bottom of the ranking. The size of the overall gap (18 points between the easiest and hardest domain) is far too large to be noise.

3. Time of day: a small afternoon advantage

For 2,968 tests we could convert the time of the test to the taker's local time with confidence (countries with a single time zone, such as the Netherlands, Spain, Indonesia or Japan). Grouped into six parts of the day, the pattern is modest but visible.

Early-morning tests (5–8 a.m., 146 tests) averaged 113.3, afternoon tests (1–4 p.m., 692 tests) 116.4. Evening tests dipped again to 114.2. A three-point difference is small, about a fifth of a standard deviation, and the early-morning group is the smallest in the sample, so we would not build a schedule around it. But it matches what sleep researchers have found about alertness over the day: a dip after waking, a peak in the early afternoon, and a slow decline into the evening.

Weekdays versus weekends showed even less variation. Saturday had the highest average (115.9) and the weekdays were all within a third of a point of each other (114.1 to 114.4). If you have a choice, a Saturday afternoon is marginally your best slot; a Monday at 7 a.m. is marginally your worst.

4. Who takes an IQ test online?

The 3,834 tests came from 134 countries, with the largest groups in China, the United States, Spain, Indonesia, the Netherlands, Poland and India. The average score was 114.5, well above the 100 the scale is built around. That is not a sign that the internet is unusually clever. Two things play a part. The first is how the scale is built: the score is computed from difficulty-weighted correct answers with a fixed formula, not normed on a representative sample, so there is no reason for any group to average exactly 100 on it. The second is who chooses to take a test: people who expect to do well, people preparing for a Mensa or job assessment, and people who already enjoy puzzles. One in seven test-takers scored 130 or above; on the textbook normal curve (mean 100, standard deviation 15) that figure would be about one in 44.

This is the single most important caveat for every number on this page. Nothing here is a national or population statistic. It is a description of several thousand people who chose to take a free online test, and it should be read that way.

Methodology and limitations

• Sample: every completed standard IQ test on IQ-EXAMS.COM between 15 April and 6 October 2026 with a recorded completion time (3,834 tests). The site's other assessments (WAIS-style, Stanford-Binet-style, Mensa-style, CHC and MoCA-style screenings, another 3,170 completions) were excluded so that all scores come from the same 25 questions.
• Scoring: the test reports a score on a scale with a mean of 100 and a standard deviation of 15, derived from the number of correct answers. The score is computed from difficulty-weighted correct answers with a fixed formula and is not normed on a representative sample. It is an online self-administered test, not a clinically administered instrument, and its scores are not interchangeable with a professionally administered IQ assessment.
• Domains: 3,810 tests had a complete per-domain breakdown (five questions per domain). The remaining 24 were older sessions without one.
• Time of day: local time was derived from the country of the visitor's connection for 51 countries with one time zone (summer time applied where in force). Multi-zone countries such as the United States, Brazil, Russia and Australia were excluded from that analysis only.
• Speed buckets: tests with a recorded time of under 60 seconds were excluded from the domain and time-of-day analyses as incomplete; they are included in the speed chart's "under 3 minutes" bar.
• Privacy: all figures are aggregates. No names, e-mail addresses or individual scores were examined or are reported. Country is the only geographic field, and no country-level score rankings are published because the samples per country are small and self-selected.
• Statistical note: with a standard deviation of about 15, the standard error of a group average is roughly 15 divided by the square root of the group size: about 0.6 points for the largest groups and 1.4 points for the under-3-minute group. The speed and domain differences are many times larger than that; the time-of-day differences are two to three times larger.

Using this data

Journalists, teachers and researchers are welcome to quote these findings and reuse the charts with a link to this page as the source. For the underlying aggregate tables, a different cut of the data (for example by language or by month), or a comment, contact our press page. We will update this study as the sample grows and note any changes here.

Related reading: what counts as a good IQ score, how IQ tests work, and whether you can improve your IQ.