ACT Score to IQ: What the 2026-27 National Ranks Do and Do Not Say

An ACT composite of 27 is at the 89th percentile of ACT-tested graduates. Matched to the IQ scale, where the mean is 100 and the standard deviation is 15, that rank would be an IQ of about 118, but only if ACT-takers were a mirror of the whole population and the ACT measured intelligence without error. Neither is true, so there is no official ACT-to-IQ conversion, and every table that prints one is making assumptions. ACT's new national ranks took effect for tests taken from September 2026, which makes this a good moment to do the arithmetic with the official numbers. This guide sets out what the ranks say, how to read them on the IQ scale, why that is not a conversion, and why the national average fell from 20.8 to 19.4.

What do ACT's 2026-27 national ranks say?

ACT's table reports a national rank for every score from 1 to 36. A rank is a percentile: the percent of ACT-tested graduates who scored at or below that score. The edition for the 2026-27 reporting year, September 2026 through August 2027, is based on ACT-tested graduates of 2024, 2025 and 2026; the edition before it used the classes of 2023, 2024 and 2025. In the new edition the composite has a mean of 19.1 and a standard deviation of 6.1, against 19.2 and 6.1 before. Nine composite scores gained one rank point and none lost any: 11, 12, 14, 16, 17, 19, 23, 27 and 29. A 27 moved from the 88th percentile to the 89th.

• Composite 36: rank 100
• Composite 34: rank 99
• Composite 33: rank 98
• Composite 30: rank 94
• Composite 27: rank 89
• Composite 24: rank 80
• Composite 21: rank 68
• Composite 19: rank 58
• Composite 18: rank 52
• Composite 15: rank 34

Two details matter for reading the table. The mean sits above the middle of the distribution: in the class of 2025 the median composite was 18 against a mean of 19.4, so the typical ACT-taker scored a little below the headline average. And since September 2025 the composite has been based only on the English, math and reading scores; science is optional and feeds a separate STEM score. ACT says its research shows consistent score distributions between the old four-section composite and the new three-section one. We found no note in the rank table saying how composites from earlier test dates were treated.

What is an ACT score on the IQ scale?

A percentile can be placed on any scale whose shape is known. On the IQ scale a rank of 50 is an IQ of 100, a rank of 84 is 115 (one standard deviation above the mean) and a rank of 97.7 is 130. Matching each ACT rank to the IQ scale this way gives the figures below. It is the same arithmetic our IQ percentile calculator and IQ score converter run.

• Composite 36 (rank 100; 99.79% of the class of 2025 scored lower): about 143
• Composite 34 (rank 99): 135
• Composite 33 (rank 98): 131
• Composite 32 (rank 97): 128
• Composite 30 (rank 94): 123
• Composite 28 (rank 91): 120
• Composite 27 (rank 89): 118
• Composite 25 (rank 83): 114
• Composite 24 (rank 80): 113
• Composite 22 (rank 72): 109
• Composite 20 (rank 63): 105
• Composite 19 (rank 58): 103
• Composite 18 (rank 52): 101

Read the top row with care. ACT's rank counts everyone at or below a score, so a 36 is rank 100 by definition. ACT's profile report shows that 2,958 of the 1,380,130 graduates in the class of 2025 earned a 36 (0.21%, about 1 in 470), so 99.79% scored lower, which is an IQ-scale value of about 143. Because each rank includes the whole of its score, every figure in the list is the top of that score's band; the middle of the band would be about one IQ point lower. The first composite to reach 130 on this arithmetic is a 33, because 130 is the 97.7th percentile and a 32 is only at rank 97. One composite point is worth about 2.5 IQ-scale points at ACT's standard deviation of 6.1 (15 divided by 6.1), but the ranks move unevenly: near a 19 a point changes the rank by 5 or 6, near a 27 by only 2 or 3, because the distribution thins out at the top.

Why is that not an ACT-to-IQ conversion?

The first problem is who sits the test. ACT states that its tested graduating class is not a nationally representative sample of graduates. About 36% of the class of 2025, 1,380,130 graduates, took the ACT, down from 59% of the class of 2015 in ACT's own 2015 report, and the share by state ran from 1% in Maine to 100% in several states where every student sits it during the school day. ACT also reports that states testing a larger share of their graduates tend to have lower mean composites, because their takers are closer to everyone, while states testing few are dominated by students who chose to test for college admission. A rank among takers is therefore a rank within a group whose make-up differs by state and by year. Whether that makes a score look better or worse than it would among all young people depends on who is missing, and some of the missing sat a different admissions test or none, so the direction cannot be read from ACT's data.

The second problem is that the ACT is not a perfect measure of intelligence. In Koenig, Frey and Detterman (Intelligence, 2008), a general-ability factor taken from the Armed Services Vocational Aptitude Battery for 1,075 participants in the National Longitudinal Survey of Youth 1979 correlated .77 with their ACT scores. In a later sample of 149 people, the correlation between the ACT composite and IQ scores derived from Raven's Advanced Progressive Matrices was .61. The same group's earlier study of the SAT (Frey and Detterman, Psychological Science, 2004) found .82 in 917 people. The ACT equation in the 2008 paper had a standard error of about 7 IQ points, and it was fitted to the pre-1990 version of the test; a 2025 paper by Jon Butler in Intelligence says that it gives less accurate estimates with newer ACT scores.

When a correlation is below 1, the best guess for a person's IQ sits closer to 100 than their test score does, a pull called regression to the mean. With r = .77 the expected IQ is 100 plus 15 times .77 times the rank's z-score (the number of standard deviations from the mean), with a spread of about 9.6 points around it. For an ACT composite that gives the following, with the 95% range in brackets.

• Composite 36: expected IQ about 133 (114 to 152)
• Composite 33: about 124 (105 to 142)
• Composite 30: about 118 (99 to 137)
• Composite 27: about 114 (95 to 133)
• Composite 24: about 110 (91 to 128)
• Composite 19: about 102 (84 to 121)

Popular conversion charts use the percentile mapping of the previous section. One widely shared chart lists a 36 at 141 to 145 and a 27 at 113 to 115, describes its method as percentile mapping adjusted for the college-bound population, and states its own margin of error as 8 to 12 IQ points. A 145 beside a 36 is a rank, not a measurement, and the correlations above say that a perfect composite predicts an IQ nearer 133. Ranking among takers moves the answer by an unknown amount in an unknown direction, and an imperfect correlation pulls the expected IQ toward 100. Neither is a reason to ignore the ACT; both are reasons to distrust a single number printed beside an ACT score. For the same reasoning applied to other exams see our guides to the SAT, the GRE and LSAT and the ASVAB and AFQT, and our wider comparison of achievement tests and IQ tests.

Why did the national average fall from 20.8 to 19.4?

ACT's average composite for the graduating class fell from 20.8 in 2016 to 19.4 in 2025. ACT's own explanation is about who sits the test. Each year it splits the class by how a student most recently tested: during the school day, in a state or district programme, or on a weekend date by choice. The table in its guide to interpreting graduating-class data gives the means and shares.

• 2016: school-day 18.3 x 34% + weekend 22.1 x 66% = 20.8
• 2019: 18.1 x 40% + 22.4 x 60% = 20.7
• 2022: 17.9 x 59% + 22.6 x 41% = 19.8
• 2025: 17.8 x 69% + 23.0 x 31% = 19.4

Our arithmetic on those figures: applying 2016's group means to 2025's mix of testers gives 19.5, so about 1.3 of the 1.4-point fall comes from the change in who tests and about 0.07 from the groups' own scores. Put the other way, had the class of 2025 kept the 2016 mix, its average would have been 21.2, which is 0.4 above 2016. Weekend testers' mean rose from 22.1 to 23.0 while school-day testers' fell from 18.3 to 17.8. That does not show that students became more or less able. It shows that a headline average moves when the mix does, the same lesson as the 270-point gap between groups of takers in our SAT class of 2026 report.

Does American Mensa accept ACT scores?

No ACT score earned today can be submitted. American Mensa's list of accepted past tests takes an ACT composite of 29 only if the test was taken before September 1989, and marks an ACT taken after that date as not accepted. The SAT is treated the same way: 1300 before 30 September 1974, 1250 up to 31 January 1994, and not accepted after. For the routes that do qualify, see our guide to Mensa's IQ requirement.

What should students and families take from the ranks?

• Read your score against the current table: a 27 is above 89% of ACT-tested graduates, a 21 above 68% and a 19 above 58%
• Remember that the rank is among ACT-takers, and that a state testing every student and one testing 1% are not comparable
• Do not turn a rank into an IQ with a formula or a table; use the IQ scale only to see the size of a gap, at about 2.5 IQ-scale points per composite point
• Treat one or two points as noise, since every test score carries a margin of error, explained in our guide to the IQ score margin of error
• If you want a result on the standard IQ scale, take a test built for that

Our IQ Metrics test reports on the scale used throughout this guide, and the IQ bell curve shows where a result sits. For how elite-university averages are inferred from admissions scores, see our analysis of the average IQ of MIT, Oxford and Yale students, and for the PSAT scale behind National Merit see National Merit 2027 cutoffs.

Common questions