What IQ is the top 5 percent, and what a rank among takers hides
Top 5 percent IQ means about 124.7 or higher on the standard scale, and top 1 percent about 134.9, when the comparison group is adults in general. Many online results instead rank you among the people who took that site's own test. This page converts the cutoffs, explains why the two ranks differ, compares the real routes to a percentile, and shows how to check which kind of number a result gives.
On a scale with a mean of 100 and a standard deviation of 15, the top 5 percent begins at about 124.7 (our arithmetic).
0 The short answer
The top 5 percent starts at an IQ of about 124.7 and the top 1 percent at about 134.9 on the standard scale (mean 100, standard deviation 15), but only when the comparison group is adults in general, and a result that says you fall between the top 1st and 5th percentiles among IQ test takers ranks you among the people who took that site's test, which is not a normed IQ. The real routes to a percentile against a defined adult group are a psychologist giving a Wechsler battery such as the WAIS-5, a supervised society test such as American Mensa's, whose qualifying score is at or above 98 percent of the general population, and a normed online battery such as ACIS, which reports a Full Scale IQ with a percentile and a 95 percent confidence interval for 15, 30 or 50 dollars one time (prices read on October 29, 2026) in about 45, 110 or 175 minutes. The decisive limit is precision: a score near 125 is a range and not a point, so read it as a place to look, and ACIS is online, unsupervised, not clinical and in English only.
124.7
IQ at the 95th percentile, the top 5 percent, on a scale with mean 100 and standard deviation 15 (our arithmetic on the normal curve). The top 1 percent starts at about 134.9.
99 versus 89
Percent of GRE takers who scored lower than a scaled score of 170 in Verbal Reasoning and in Quantitative Reasoning, July 1, 2022 to June 30, 2025 (ETS): one top score, two ranks, because the takers differ.
115 to 130
The 95 percent confidence interval printed beside a verbal composite of 124, ranked at the 95th percentile, in Pearson's sample WAIS-5 report.
1 What IQ Is the Top 5 Percent, and What IQ Is the Top 1 Percent?
On a scale with a mean of 100 and a standard deviation of 15, the top 5 percent begins at an IQ of about 124.7 and the top 1 percent at about 134.9, and these cutoffs are our own arithmetic on the normal curve, not a finding of any study. The method is short. An IQ score on this scale is a standard score, so a person at 100 is at the mean and each 15 points is one standard deviation. The NIST/SEMATECH e-Handbook of Statistical Methods, which we opened on October 29, 2026, defines the normal distribution by a location parameter, the mean, and a scale parameter, the standard deviation, and calls the case with a mean of 0 and a standard deviation of 1 the standard normal distribution. The 95th percentile of that distribution sits 1.645 standard deviations above the mean, which is 100 plus 1.645 times 15, or 124.7. The 99th percentile sits 2.326 standard deviations above it, which is 134.9.
Top share
Percentile
Standard deviations above the mean
IQ cutoff on the SD 15 scale (our arithmetic)
People per 100 adults at or above it
Top 10 percent
90th
1.282
119.2
10
Top 5 percent
95th
1.645
124.7
5
Top 2 percent
98th
2.054
130.8
2
Top 1 percent
99th
2.326
134.9
1
Rounded to whole scores, an IQ of 125 is at the 95.2nd percentile and an IQ of 135 at the 99.0th, which is how the page on the IQ percentile chart prints them. The full inverse table, from any percentile back to a score, is on the page about converting a percentile to an IQ, and the page on how rare a given IQ is restates the same cut as one person in a number. "Top 5 percent" and "95th percentile" name the same cut from opposite ends: the top share equals 100 minus the percentile. The glossary of the Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014), which we opened in its open access edition, defines a percentile as the score below which a given percentage of scores for a specified population occurs. The words "a specified population" carry the rest of this page.
Two limits come with the table. The first is the scale. A scale with a standard deviation of 16 puts the top 5 percent at 126.3 and one with a standard deviation of 24 at 139.5 (our arithmetic), so a published cutoff only transfers if the standard deviation matches, and the page on why the standard deviation is 15 explains which scales use which. The second is the group. The table assumes that scores follow a normal curve in a population of adults, so it answers "how far above the average adult is this score?" It does not say where a person ranks among those who took a given website's test, and it cannot say where they rank in the world. The norm groups described in the sources we opened are national or general population samples. Pearson's U.S. brochure for the WAIS-5, for example, says the normative samples were collected in 2023 and 2024 and match the most current census on five key demographic variables. That is a national sample, not a sample of everyone alive, and we found no IQ norm in the sources we opened that states a world reference group. A person who asks about "the top 5 percent in the world" is therefore asking a question no norm table we read was built to answer, and the practical translation is the top 5 percent of the adults the test was normed on.
The same top 5 percent cutoff reads differently on other scales. Abbie Salny, American Mensa's supervisory psychologist, writes in The Question of Genius that the Cattell IIIB has a mean of 100 and a standard deviation of 24, against 16 for the Stanford-Binet and many school tests. The sourced account of the Cattell standard deviation also shows how to convert any such score to the 15 point scale.
For a sense of scale, 5 in 100 adults is 1 in 20, and the top 1 percent is 1 in 100, again our arithmetic. The page on what a good IQ is puts these cutoffs beside the common descriptive bands.
A rank always depends on who is in the comparison group. Pearson says the ideal norm group represents those taking the test in the context at hand, so an employer might compare applicants with peers in the profession, while someone interested in intelligence testing might compare with the general population. The guide to choosing and reading Watson-Glaser norm groups shows how that choice moves the number.
2 What Does "Your Score Falls Between the Top 1st and 5th Percentiles Among IQ Test Takers" Mean?
That sentence reports a rank inside a group of people who took one test, and because it gives a band instead of a score it is a percentile statement and not an IQ, so it cannot be converted into the 124.7 or 134.9 above unless the group of test takers happens to match the adult population. We could not identify which website prints this exact line: our searches on October 29, 2026 returned calculators, tables and articles, not a result page with that wording. So this section reads the sentence and not a product, and it says nothing about any company's test. The sentence has four parts, and each deserves a careful reading.
"Top 1st to 5th percentiles" is a band, not a point. Read at face value, "top 5th percentile" means the highest 5 percent of the group and "top 1st percentile" the highest 1 percent, so the band describes someone ahead of between 95 and 99 percent of the group. The wording is also ambiguous in a way that matters. "First percentile" in a percentile table normally means the bottom 1 percent, so a reader who sees "1st and 5th percentiles" without the word "top" could take the sentence the wrong way round. The Standards define a percentile rank as the rank of a score based on the percentage of scores in a specified distribution that are below it, so the direction is always set by the definition, and a report should print it.
"Among IQ test takers" names a group, and it does not say which. It could mean everyone who ever took that site's test, everyone who took it that month, a norm sample the publisher recruited, or the people who finished all items. Each choice produces a different rank for the same performance. The Standards set the expectation in Standard 5.8: norms, if used, should refer to clearly described populations, and the populations should include the groups with whom test users will ordinarily wish to compare their examinees. A sentence that names "IQ test takers" without describing them does not meet that expectation by itself.
The sentence says nothing about the score. A percentile rank is a position, and an IQ is a score on a scale whose mean is fixed at 100. Two people can both be in "the top 5 percent among takers" with very different performance if the takers differ, and the same performance can fall in the top 5 percent of one site's takers and at the 80th percentile of another's. The page on why you cannot average scores and percentiles explains the difference between the two numbers, and the IQ percentile calculator converts a score you already hold from a normed test.
The sentence also does not say how precise the band is. A band of "1st to 5th" spans four percentile points at the top of the distribution, which on the SD 15 scale is about 10 points of IQ (124.7 to 134.9, our arithmetic) if the group were all adults. That range is wide, and the section on precision below shows that a single score can carry an interval of about 15 points. Speed does not fix any of this. The page on IQ tests that give instant results covers what an immediate score does and does not tell you, and the answer starts with the same question: against whom?
Employer percentiles carry a precision question as well. The Standards say that if an assessment is shortened the reliability is likely to decrease, and SHL's sample report says that because each candidate receives a different set of items there is no direct correlation between the share answered correctly and the percentile achieved. The explanation of why a numerical percentile is an estimate sets this out.
3 Why a Rank Among Test Takers Is Not a Normed IQ
A percentile among a test's own takers is a real statistic, but it describes a self assembled group, so it measures how you compare with those people and not how you compare with adults in general, and the same performance can earn very different ranks in different groups. The clearest published example comes from graduate admissions, where the reference group is stated. The Educational Testing Service publishes the GRE Guide to the Use of Scores (2026 to 2027 edition), which we opened on October 29, 2026. Its tables give, for each scaled score, the percent of test takers who scored lower, based on all individuals who tested between July 1, 2022 and June 30, 2025, and it reports 788,021 takers on the Verbal Reasoning measure and 790,864 on Quantitative Reasoning over that period.
The same top scaled score earns two ranks. A Verbal Reasoning score of 170 was higher than the scores of 99 percent of takers, and a Quantitative Reasoning score of 170 was higher than the scores of 89 percent. At a scaled score of 160 the two measures give 82 and 50. The scales are the same, the program is the same, and the same number sits at very different ranks, because the distribution of scores among takers differs between the two measures. ETS also warns users that percentile ranks for GRE tests may change over time because they always rest on the takers within a given three year period, and that Subject Test percentile ranks should only be compared within the same Subject Test because each rests on a different reference population. Its example is that a 79th percentile rank on the Physics Test is not equivalent to a 79th percentile on the Mathematics Test.
That is the model for what any "among test takers" line is: a legitimate number, defined by a named group and a date range, and useful for the decision it was designed for, which here is comparing applicants. Turning a GRE percentile into an adult IQ would be a mistake, and so is turning a website's rank into one. A rank among those who chose to take an online test is a rank among people who chose to, and the choice shapes the group, as the next section shows.
Hiring tests supply another example. Criteria Corp reports the CCAT as a raw score out of 50 plus a percentile against a norming group of 106,740 adults aged 18 and older, incumbents and applicants, whose median raw score is 24. The explanation of the CCAT raw score and percentile notes that this is the middle of people tested for work, not of adults in general.
The Standards give the general rule in their chapter on scores and norms. They state that the validity of norm referenced interpretations depends in part on the appropriateness of the reference group, that it is important that reference populations be carefully defined and clearly described, and that when only a sample of the reference population is tested it is important that the norms be based on a technically sound, representative sample of sufficient size. Standard 5.3 adds that when there is sound reason to believe specific misinterpretations of a score scale are likely, users should be explicitly cautioned. A result line that invites readers to hear "top 5 percent" as "IQ 125" is a candidate for that caution. The page on how IQ scores are normed walks through the chain that turns a raw score into a standard score against a reference sample, and it is the right companion to this section.
None of this says a site is dishonest. A rank among takers can be computed correctly and described honestly. The problem is that a reader searching "top 5 percent IQ" wants the other number, the one against adults, and a rank among takers cannot supply it without a documented link between the two groups.
4 How Much Can Self Selection Move a Rank Among Test Takers?
Because people who take an online IQ test chose to, and the choice can depend on interest, confidence or prior scores, a rank among those takers can differ from a rank among adults by many percentile points, and the direction depends on who shows up. The general principle is documented for web surveys. Jelke Bethlehem's review of selection bias in web surveys, published in the International Statistical Review in 2010, states that specific groups are under represented because they have less access to the Internet, that recruitment of respondents is often based on self selection, and that both under coverage and self selection may lead to biased estimates. He concludes that under coverage problems may solve themselves in the future, but that self selection leads to unreliable survey outcomes. His subject is surveys, not IQ tests, and we cite him for the mechanism, not for any estimate of how selected IQ test takers are.
The mechanism can be shown with invented numbers. Suppose the takers of some site have a true mean IQ of 100, 105, 110 or 115 against adult norms, with a standard deviation of 15 in each case. These are hypothetical groups, not measurements of any site, and every figure below is our arithmetic on the normal curve.
Hypothetical mean IQ of the site's takers
Rank among takers of a person whose adult normed IQ is 125
IQ needed to be in the top 5 percent among those takers
IQ needed to be in the top 1 percent among those takers
100 (same as adults)
95th percentile
124.7
134.9
105
91st percentile
129.7
139.9
110
84th percentile
134.7
144.9
115
75th percentile
139.7
149.9
The table runs in both directions. If takers are more able than adults in general, a person with an adult normed IQ of 125 can land at the 84th percentile among takers, and "the top 5 percent among takers" can require a normed IQ near 135. If takers are less able, the same person can look more distinguished. The shift in the band is the shift in the takers' mean, and since a site rarely publishes who its takers are, a reader cannot undo it. The band "top 1st to 5th percentile" would correspond, in the hypothetical group with a mean of 110, to adult normed scores of roughly 135 to 145. In the group with a mean of 100 it corresponds to 125 to 135. Nothing on the result line says which case applies.
There are further effects that the table leaves out. People who take an online test twice, or who try several tests and keep the result they prefer, can change the group's composition, and the page on retaking an IQ test covers what a second sitting does to a score. A pool of takers can also drift over time: a site that draws students one month and professionals the next has a different distribution in each, so a rank computed on an accumulating pool can move even if the test does not. The page on whether online IQ tests are accurate takes up the wider question of what to trust. For this page the point is narrow: unless a site documents who is in its group, a rank among its takers is a fact about the site, not about you.
5 Which Routes Give a Percentile Against Adults, and What Does Each Cost?
Only a test with a described adult reference group gives a percentile you can read as a position among adults, and the real routes differ in whether a person supervises you, how long they take, what they cost and what you receive. The table compares them. Prices and facts are those we read on October 29, 2026, and they can change.
Option
What it measures
Normed against whom
Supervised
Time
Cost, read October 29, 2026
What you receive
Psychologist giving a Wechsler battery (WAIS-5)
Five primary indices and a Full Scale IQ
Pearson's U.S. norms, collected 2023 to 2024 and matched to the most current census on five demographic variables
Yes, an examiner administers it
Not stated in the sources we opened
No single published price; it depends on the clinic
A written report; the sample report prints a percentile rank and a 95 percent confidence interval for each composite
American Mensa proctored test
A qualifying score at 98 percent of the general population
The general population, per Mensa's testing page
Yes, a proctor, in a local group session or through private electronic testing
Not stated on the page we opened
60 dollars for local group testing and 99 dollars for private testing (regular prices)
A qualifying or non qualifying result for membership
Free online screener or game
Varies; a short sample of items
Often not stated, or the site's own takers
No
Minutes
Free
One number or a band, often a rank among the site's takers
Practice test built for job screening
Reasoning items under time pressure
Employer or vendor groups
Usually no
Minutes
Varies
A score or band for a screening decision; it is not an IQ test
ACIS Quick
6 subtests in 3 domains
Adult norms covering ages 16 to 90
No
About 45 minutes
15 dollars one time
Full Scale IQ with percentile and 95 percent confidence interval
ACIS Optimized
13 subtests in 5 domains
Adult norms covering ages 16 to 90
No
About 110 minutes
30 dollars one time
Full Scale IQ with percentile and 95 percent confidence interval
ACIS Full Scale
All 20 subtests in 6 domains
Adult norms covering ages 16 to 90
No
About 175 minutes
50 dollars one time
Full Scale IQ, six indices, percentile and 95 percent confidence interval
Read the table by what you need. If a result must be defensible to a third party, for example for a clinical question, the first row is the route, and the page on every way to take an IQ test covers the process. If you want membership in a society, the second row is the route, and a later section explains how its line compares with the top 5 percent. If you only want to see how a normed test feels before deciding, a free IQ test is a fair start, and the page on free versus validated IQ tests explains what separates the two kinds of result. If you want a normed percentile for yourself without a clinic visit, the page on the best online IQ tests compares the normed batteries, and the IQ test for adults describes how ACIS fits that reader.
The free and screening rows deserve a plain reading. A short screener can produce a sound number only if its items are numerous enough and its norms are described, and a job screening practice test is built to prepare for a screening decision, not to place a person among adults. Neither is a defect. They answer different questions. What matters for the top 5 percent question is the column headed "Normed against whom": if the cell says nothing, or says the site's own takers, the percentile is a rank among takers.
An examiner given screener shows the other side of that rule. The Shipley-2 takes 20 to 25 minutes, reports standard scores, percentiles, age equivalents and confidence intervals against a norm sample of 2,826 individuals, and is sold to qualified professionals, yet the WPS and PAR pages do not give the dates of data collection. The page on the Shipley-2 and its norm sample examines what that leaves open.
The ACIS rows come from the home page, read on October 29, 2026. Each form is a one time payment with no subscription, and the Custom Test, at 5 dollars per subtest, returns the full index of any domain completed.
6 How Can You Tell Whether a Percentile Comes From a Defined Adult Group? Five Checks
A percentile can be read as a position among adults when five questions have written answers: who is in the reference group, how they were recruited, when they were tested, which ages the norms cover, and whether the report prints an interval around the score. These are not our invention. Standard 5.9 of the Standards says that reports of norming studies should include a precise specification of the population that was sampled, the sampling procedures and participation rates, any weighting of the sample, the dates of testing and descriptive statistics, and that technical documentation should indicate the precision of the norms themselves. The comment adds that the information should be sufficient to enable users to judge the appropriateness of the norms for interpreting the scores of local examinees. The APA Guidelines for Psychological Assessment and Evaluation, approved by the APA Council of Representatives in March 2020, make the same point from the user's side. Guideline 11 asks psychologists to understand the test manual information on the characteristics of the norm sample, and its discussion warns that normative data may not be accurate when important features of the examinee are not represented in the norm group.
Check
The question to ask
What a documented answer looks like
1. Population
Who is in the reference group?
Pearson's U.S. WAIS-5 brochure: samples that match the most current census on five key demographic variables. Mensa: the general population
2. Recruitment
How were they chosen, and how many agreed?
Standard 5.9 asks for the sampling procedures and the participation rates. A sentence that says "IQ test takers" gives neither
3. Dates
When were they tested?
Pearson: 2023 and 2024. ETS: July 1, 2022 to June 30, 2025 for its GRE tables
4. Ages
Which ages do the norms cover?
ACIS: adult norms covering ages 16 to 90, per its home page on October 29, 2026
5. Interval
Does the report print precision?
Pearson's sample report prints a percentile rank and a 95 percent confidence interval for each composite. ACIS reports a Full Scale IQ with its 95 percent confidence interval
Run any "top 5 percent" result through the table. A result that is silent on checks 1 to 3 is a rank among whoever happened to take the test. That does not make the rank false, as the section on ranks among takers showed, and it does make it a different number from the one most readers are looking for. A result that answers all five is a norm referenced score, and its percentile can be read as a position among the reference adults, to the precision the interval shows.
The documentation that answers the five questions for ACIS sits in the technical manual, which a buyer can read before paying and which we do not quote here: the page on the ACIS technical manual is the place to read how its norms and scores are documented. The page on what makes an IQ test scientific sets out the wider checklist that these five checks belong to. One more point belongs here. The Standards call for a reference group "with whom test users will ordinarily wish to compare" examinees (Standard 5.8), and a reader who wants to know whether they are in the top 5 percent of adults wants the adult group. If a page does not say that its group is adults, assume it is not.
7 How Precise Is a Score at the Top 5 Percent Line?
A score near 125 is a range and not a point, because every score carries measurement error, so whether a person is in the top 5 percent can depend on an interval that spans from the top 16 percent to the top 2 percent.Pearson's sample WAIS-5 score report, which we opened on October 29, 2026, shows the shape. For its sample examinee, a woman of 70 tested in 2024, the verbal comprehension composite is 124 with a percentile rank of 95 and a 95 percent confidence interval of 115 to 130. Converting the ends of that interval with the normal curve (our arithmetic), 115 is at the 84th percentile and 130 at the 98th. So a composite that the report ranks at the 95th percentile has an interval that runs from the top 16 percent to the top 2 percent. The same report prints a Full Scale IQ of 111, at the 77th percentile, with an interval of 106 to 116, which in our arithmetic spans the 66th to the 86th percentile. The report states that its confidence intervals are calculated using the Standard Error of Estimation.
Three things follow. The first is that a single point is fragile at the line. In our arithmetic an IQ of 124 is at the 94.5th percentile and an IQ of 125 at the 95.2nd, so a one point difference moves a person across the top 5 percent boundary while sitting well inside any plausible interval. The second is that the interval is wider for a narrower score. In the sample report the Full Scale interval spans 10 points and the verbal composite interval 15, which fits the general rule that a score built from more tasks is more precise. The Standards say so in Standard 2.3: estimates of reliability and precision should be reported for each total score, subscore or combination of scores that is to be interpreted, and the comment adds that reporting them only for total scores is not sufficient when subscores are also interpreted. The third is that precision varies by level. Standard 2.14 asks that conditional standard errors of measurement be reported at several score levels unless there is evidence that the standard error is constant, and that where cut scores are specified, the standard errors be reported in the vicinity of each cut score. The top 5 percent line is such a cut, and a reader who is near it should ask for the precision at that level, not only the average.
The practical rule is simple. As our own rule of thumb, not Pearson's, treat a result within about five points of 125 as consistent with being on either side of the top 5 percent line, and read the interval before the point. The page on reliability and validity explains why a score can be consistent without being meaningful, and the page on the IQ score chart places each score band beside its percentile. A second sitting does not settle a close call either: the page on what a second sitting changes explains that a new result is another draw around the same underlying level, not a verdict.
What our arithmetic does and does not showThe percentile range for the sample composite converts the ends of a printed interval with the normal curve. It does not come from Pearson, it ignores any adjustment a publisher applies, and the sample report belongs to an invented examinee. We use it only to show how wide a 95 percent interval looks when expressed as ranks.
8 Top 5 Percent of What: A Full Scale IQ, One Index or One Subtest?
A percentile can be reported for a total, for a broad ability index or for a single subtest, and a person can be in the top 5 percent on one without being near it on another, so the first question is which score the percentile belongs to. Pearson's sample report makes the point. Its examinee's verbal comprehension composite is 124 at the 95th percentile, her fluid reasoning composite is 105 at the 63rd, her visual spatial composite 103 at the 58th, her working memory composite 97 at the 42nd, her processing speed composite 108 at the 70th and her Full Scale IQ 111 at the 77th. One index sits at the top 5 percent mark while the total does not. That pattern is not unusual for a profile, and the page on index score differences covers how common large gaps between indices are in ordinary people and how to read them.
The Cattell-Horn-Carroll (CHC) model names the abilities behind the indices, and the page on the CHC model explains where the names come from. ACIS reports six: verbal comprehension (VCI, crystallized knowledge, Gc), fluid reasoning (FRI, Gf), quantitative reasoning (QRI, Gq), visual spatial (VSI, Gv), working memory (WMI, Gwm) and processing speed (PSI, Gs). The page on the cognitive domains in IQ testing describes each broad ability, and the ACIS pages for each subtest describe its task format. A top 5 percent result on the Full Scale IQ asks for strong performance across the subtests that build it. A top 5 percent result on one index asks for it in one ability, and that is easier to reach by chance when you look at six indices than when you look at one total, because every extra index is another chance for ordinary variation to land in the tail. That is our reasoning, not a finding.
What a percentile means also depends on the form. Every ACIS form reports a Full Scale IQ with its 95 percent confidence interval, and the Quick form samples 3 domains, the Optimized form 5 and the Full Scale form all 6. If you want a percentile for one ability only, the Custom Test returns the full index for any domain you complete. For speed in particular, the page on processing speed testing explains what a speed index measures. A percentile for a single subtest is the least precise of the three, for the reasons given in the previous section.
9 Does the Age of the Norms Change Who Is in the Top 5 Percent?
It can, because IQ test scores have risen over time, so a score earned against old norms can be higher than the same performance would earn against current norms, and a top 5 percent claim should say when the reference group was tested. The best known estimate comes from a meta-analysis. Lisa Trahan, Karla Stuebing, Jack Fletcher and Merrill Hiscock, in a 2014 article in Psychological Bulletin whose abstract we read through PubMed on October 29, 2026, report that across 285 studies (N = 14,031) since 1951, with administrations of two intelligence tests with different normative bases, the meta-analytic mean was 2.31 standard score points per decade (95 percent confidence interval 1.99 to 2.64). For 53 comparisons (N = 3,951) involving modern Stanford-Binet and Wechsler tests, the mean was 2.93 points per decade (95 percent confidence interval 2.3 to 3.5), which the authors describe as comparable to previous estimates of about 3 points per decade and not consistent with the hypothesis that the Flynn effect is diminishing.
Applied to a case, and as our arithmetic only, 20 years of drift at 2.31 points per decade is about 4.6 points, and at 2.93 points per decade about 5.9 points. A person who scored 125 against norms 20 years old would then sit near 119 or 120 against current norms, which is about the 90th or 91st percentile and not the 95th. This is an illustration, not a correction factor: the meta-analytic mean is an average across studies, tests and people, and no single test or person is guaranteed to follow it. The Standards address the same risk. Their background section on norms states that the usefulness of norms based on a given sample may diminish over time and that renorming may be required, and Standard 5.11 makes the publisher responsible for renorming with sufficient frequency to permit continued accurate and appropriate score interpretations as long as a test remains in print.
Professional guidance says the same thing to users. Standard 9.08 of the American Psychological Association's Ethical Principles of Psychologists and Code of Conduct (effective January 1, 2017), which we read in the published code, says that psychologists do not base decisions on data or test results that are outdated for the current purpose or on tests that are obsolete. Guideline 8 of the 2020 APA assessment guidelines lists the agedness of the norms among the considerations in test selection. For a reader the lesson is concrete: check the year the norms were collected, as Pearson prints it for the WAIS-5 (2023 to 2024), and be wary of a percentile from a test that does not say.
10 Is the Top 5 Percent a Society Seat or a Gifted Label?
A top 5 percent result does not reach the usual society line, which is higher, and it does not define giftedness either, because both lines are set by an institution and vary by test, so the percentile answers a different question from the label.American Mensa's testing page, which we opened on October 29, 2026, says that a qualifying score indicates that you have tested at or above 98 percent of the general population. Its regular prices were 60 dollars for local group testing and 99 dollars for private testing on that date. On the standard scale, the 98th percentile is an IQ of about 130.8 (our arithmetic), above the 124.7 that marks the top 5 percent. A person at the 95th percentile of adults is therefore not at the Mensa line by that definition, while a person at the 99th is above it. The page on the Mensa IQ test covers what score you need and how to estimate yours. ACIS is not for admission to high IQ societies, and its report should not be offered to one as a qualifying score.
Giftedness has its own pages. The page on gifted adults covers the signs, the assessment and the life outcomes, and the page on a gifted test for adults covers how an adult can confirm where they stand. This page sets no definition of giftedness: a percentile is an input to a decision and not the decision. For readers whose question is about the top 1 percent and above, the page on high range IQ testing covers the question of measuring in the upper tail. Whatever the label, the starting point is the same: a percentile against adults, read with its interval.
A last caution applies to people who are chasing the number. The top 5 percent is a statistical line and not a measure of worth or of achievement, and the group statistics behind any percentile are never a diagnosis or a forecast for one person. A high percentile on a short online test describes how a person performed on that test, in that session, against that reference group.
11 What Makes an Unsupervised Percentile Worth Reading?
An unsupervised session gives a percentile worth reading only if the conditions resemble the standardized ones, which means working sound, a quiet room, a full effort and honest timing, because no examiner is there to notice what went wrong. The APA assessment guidelines say that without systematically assessing effort it becomes difficult to discern whether variability and patterns of test results reflect actual performance or the influence of low effort, motivation or some factor besides ability, and they stress that the validity of responses may be compromised when standardized administration is not followed. Standard 9.06 of the APA Ethics Code asks psychologists, when interpreting assessment results, to take into account the purpose of the assessment, test factors, test taking abilities and other characteristics of the person assessed, including situational, personal, linguistic and cultural differences, and to indicate significant limitations of their interpretations. A person reading their own online result has no psychologist to do this, so the reader has to.
For ACIS the conditions are practical. Digit Span, Alphanumeric Sequencing and Arithmetic present their items as audio and need working sound, so check your speakers or headphones before you begin. Breaks are allowed, progress is saved between subtests, and the buyer has 30 days to finish, which means a long form can be split across days, and a person who feels tired or distracted can stop and return. The manual states that verbal and knowledge tasks are affected by English proficiency and educational exposure, and ACIS is in English only, so a person whose first language is not English should expect the verbal indices to reflect that.
Two habits protect the number. The first is to take the test once, honestly, and read the report before deciding to test again, because a rushed second sitting is a weaker measure than one honest first sitting. The second is to keep a note of the conditions: whether you were rested, whether the sound worked, whether you were interrupted. A percentile from a disturbed session is a smaller claim than the same percentile from a good one, and the interval cannot capture the disturbance. The page with an IQ test report example shows what a report prints so that you know what to read and what to look for when your own arrives.
12 Where ACIS Sits, and What a Buyer Should Do
ACIS is a normed online battery that returns a Full Scale IQ with a percentile and a 95 percent confidence interval against adult norms covering ages 16 to 90, and it is not a clinical or supervised instrument, so it can tell you where you stand against adults to the precision of its interval and it cannot certify a top 5 percent or top 1 percent status for anyone else. The facts are those on the ACIS home page, read on October 29, 2026, and prices can change. There are three forms, each a one time payment with no subscription: Quick at 15 dollars (6 subtests in 3 domains, about 45 minutes), Optimized at 30 dollars (13 subtests in 5 domains, about 110 minutes) and Full Scale at 50 dollars (all 20 subtests in 6 domains, about 175 minutes). The Custom Test costs 5 dollars per subtest, and at 10 subtests the Full Scale form costs the same. The free trial needs no card, and there is one free trial per person. The guarantee reads: "Full refund within 5 days if the report does not deliver the features described at checkout or a technical issue prevents you from accessing your results."
The limits are plain. The test is online and unsupervised, so the conditions in the previous section are the reader's responsibility. It is not a clinical or diagnostic instrument, and its results are not for hiring, school accommodations or admission to high IQ societies. It is in English only. Its report is a description of performance on its tasks against its reference adults, and the interval is part of the result, not a footnote. The reasoning is that of the Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014): a score has to be interpreted in light of the reference group, the precision and the purpose, and, as Standard 1.14 puts it for subscores and profiles, the rationale and evidence must pertain to the specific score to be interpreted. APA Guideline 8 and Ethics Code Standard 9.06 point the same way, toward the appropriate normative comparison and an honest statement of limits.
What a buyer should do depends on the question. If you only saw a line such as "top 1st to 5th percentiles among IQ test takers" and want to know whether it means an IQ near 125 to 135, the answer is that it does not tell you, and the way to find out is to take a test with described adult norms. If you want a percentile against adults, open the example report first, because it shows what each form prints, then start the free trial, which needs no card, and choose the form that covers the domains you care about. If you are near the top 5 percent line, expect the interval to straddle it and decide in advance what you would do with either side of it. If you need a result that a third party will rely on, book a psychologist.
Every third party fact on this page comes from one of the documents below, opened or checked on October 29, 2026, and every figure we calculated ourselves is labeled as our arithmetic.
American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing. AERA, 2014. Glossary entries for percentile, percentile rank and norms; the chapter introduction on norms; Standards 1.14, 2.3, 2.14, 5.3, 5.8, 5.9 and 5.11. Open access edition.
American Psychological Association. Ethical Principles of Psychologists and Code of Conduct. Effective January 1, 2017. Standards 9.06, Interpreting Assessment Results, and 9.08, Obsolete Tests and Outdated Test Results. APA.
American Psychological Association, APA Task Force on Psychological Assessment and Evaluation Guidelines. APA Guidelines for Psychological Assessment and Evaluation. Approved by the APA Council of Representatives, March 2020. Guidelines 8 and 11. APA.
Educational Testing Service. GRE Guide to the Use of Scores, 2026 to 2027. Percentile ranks for the GRE General Test, Tables 1B and 1C, based on all individuals who tested between July 1, 2022 and June 30, 2025, and the guidance on reference groups. ETS.
Pearson. WAIS-5 overview brochure (U.S. edition), section on new norms. Pearson.
Pearson. WAIS-5 sample score report for a sample examinee, report dated September 16, 2024, composite score summary. Pearson.
American Mensa. Testing page, qualifying score definition and regular testing prices, read October 29, 2026. American Mensa.
Bethlehem, J. Selection bias in web surveys. International Statistical Review, 2010, volume 78, issue 2, pages 161 to 188. DOI 10.1111/j.1751-5823.2010.00112.x.
Trahan, L. H., Stuebing, K. K., Fletcher, J. M. and Hiscock, M. The Flynn effect: A meta-analysis. Psychological Bulletin, 2014, volume 140, issue 5, pages 1332 to 1360. DOI 10.1037/a0037173.
National Institute of Standards and Technology. NIST/SEMATECH e-Handbook of Statistical Methods, section 1.3.6.6.1, Normal Distribution. NIST.
14 Frequently Asked Questions
What IQ is the top 5 percent?
The top 5 percent starts at an IQ of about 124.7 on the standard scale, where the mean is 100 and the standard deviation is 15, which is the 95th percentile. In whole numbers, 125 is the usual cutoff. That figure is our arithmetic on the normal curve and assumes the comparison group is adults.
What IQ is the top 5 percent in the world?
No IQ norm we found uses a world reference group, so the question has no direct answer. The practical reading is the top 5 percent of the adults a test was normed on, about 124.7 on a standard deviation 15 scale. A national norm sample describes one country, not everyone alive.
What is the top 5 percentile IQ?
The top 5 percentile means the 95th percentile, an IQ of about 124.7 on the standard scale. People use the phrase loosely, so check the direction. A 5th percentile, without the word top, normally means the bottom 5 percent, which is about 75.3 by our arithmetic.
What IQ is the top 1 percent?
The top 1 percent starts at about 134.9 on the standard scale, the 99th percentile, so a whole number score of 135 is the usual reading. This is our arithmetic for adults. A scale with a larger standard deviation puts the same line higher, for example 137.2 at a standard deviation of 16.
What does "your score falls between the top 1st and 5th percentiles among IQ test takers" mean?
It means the site ranked you ahead of roughly 95 to 99 percent of the people who took its own test. It does not give an IQ, and it does not say that group matches adults in general. Treat it as a rank among takers until the group is described.
Is an IQ of 130 in the top 5 percent?
Yes, against adults on a standard deviation 15 scale. An IQ of 130 is at about the 97.7th percentile, the top 2.3 percent, so it sits inside the top 5 percent and just short of the top 1 percent line at 134.9. These are our arithmetic on the normal curve.
Is the top 5 percent the same as the 95th percentile?
Yes, they name the same cut from opposite ends. If 95 percent of a group scores below a cutoff, 5 percent score at or above it, so the top share equals 100 minus the percentile. Printed ranks can differ in whether they count scores below or at or below, so read the definition.
Why can a percentile among test takers differ from a percentile among adults?
The two groups differ. A rank among takers depends on who chose to take the test, while a rank among adults depends on a documented adult sample. The GRE shows the effect: a scaled score of 170 was higher than 99 percent of Verbal takers and 89 percent of Quantitative takers from 2022 to 2025.
Does a different standard deviation change the top 5 percent cutoff?
Yes. The cutoff is the mean plus 1.645 standard deviations, so it moves with the scale. With a mean of 100 it is 124.7 at a standard deviation of 15, 126.3 at 16 and 139.5 at 24, by our arithmetic. Confirm the scale before comparing scores from different tests.
How precise is a score near the top 5 percent line?
Not very. Pearson's sample WAIS-5 report prints a verbal composite of 124, at the 95th percentile, with a 95 percent confidence interval of 115 to 130. By our arithmetic that interval covers the 84th to the 98th percentile, so a score near the line can fall on either side.
Do older norms inflate a top 5 percent score?
They can. A 2014 meta-analysis found IQ scores rising by 2.31 standard score points per decade across 285 studies, so a score earned against norms 20 years old could overstate standing against current adults by about 5 points on average, by our arithmetic. No single person is guaranteed to follow that average.
Does Mensa require the top 5 percent?
No, its line is higher. American Mensa says a qualifying score means you tested at or above 98 percent of the general population. That is about 130.8 on the standard scale by our arithmetic, above the 124.7 that marks the top 5 percent.
Can one subtest put me in the top 5 percent?
A single subtest can produce a high scaled score, but it is the least precise score a report prints, and one score among many is easy to reach by chance. A top 5 percent claim is stronger for a Full Scale IQ or an index built from several subtests, each with an interval.
Do other tests report a percentile among test takers?
Yes. The GRE program reports the percent of test takers who scored lower, based on everyone who tested in a stated three year window. That is a legitimate number with a named group, and ETS warns that percentile ranks on different tests rest on different reference populations and are not interchangeable.
How can I get a percentile against adults?
Take a test whose reference group is described as adults, such as a psychologist administered Wechsler battery or a normed online battery, and read the percentile with its confidence interval. Check the year of the norms, the ages covered and whether the report prints precision. ACIS reports against adult norms covering ages 16 to 90.
How do I tell whether a site's percentile is among adults or takers?
Look for the reference group. If the result says "among test takers" or names no group, no dates and no sampling, treat it as a rank among takers. A norm referenced result says who the comparison adults are, when they were tested and which ages the norms cover.
Does ACIS report a percentile?
Yes. Every ACIS form reports a Full Scale IQ with a percentile and a 95 percent confidence interval against adult norms covering ages 16 to 90, and the Full Scale form adds six indices. ACIS is online, unsupervised, not clinical and in English only, so it is not a certified credential.
Can an online test confirm I am in the top 1 percent?
It can estimate where you stand, with an interval, but it cannot certify membership of any group. At the top 1 percent line of about 134.9, an interval of several points on each side would straddle the line. For a result that others will rely on, use a psychologist.
Should I retake a test to reach the top 5 percent?
Retaking to chase a number is a weak plan, because a second result is another draw around the same underlying level, so a close call usually stays a close call. Retake only if the first session was disturbed, and write down the conditions so the two results can be compared.
Does the top 5 percent mean I am gifted?
Not by itself. Giftedness is a label set by a definition or an institution, so a percentile is an input and not a verdict. A top 5 percent result describes performance on one test against one reference group, and group statistics are never a diagnosis of an individual.
What should I do with a result near the cutoff?
Read the interval before the point estimate, note the test conditions, and decide in advance what each side of the line would change. A score within a few points of 125 can sit on either side of the top 5 percent, so avoid claiming the rank publicly on a single number.
Take the assessment
You get a profile, not a number
ACIS measures six CHC domains across 20 subtests and reports each one with its own normed score and confidence interval, so you can see where you are strong and where you are not.