The AQ-10 Test: The Two-Minute Autism Screener (NHS)
Ten questions, a cutoff of six, and a place in national clinical guidance. It is also the instrument that performed no better than chance in the largest study to test it against real diagnostic outcomes.
The AQ-10 Test exists for one job: to help a busy clinician decide, in under two minutes, whether an adult sitting in front of them should be referred for a full autism assessment. It was never intended to tell anyone whether they are autistic.
Almost everywhere you encounter it online, it is being used for exactly that. Understanding the gap between those two uses is the whole point of this page.
What is the AQ-10 Test?
The AQ-10 Test was published in 2012 by Carrie Allison, Bonnie Auyeung and Simon Baron-Cohen at the Autism Research Centre in Cambridge. They took the fifty items of the full AQ-50 Test and looked for the smallest subset that preserved most of the discriminating power.
Ten items survived. Just ten. Two from each of the five original subscales: social skill, attention switching, attention to detail, communication, and imagination.
You answer each on a four-point agreement scale. Definitely agree, slightly agree, slightly disagree, definitely disagree. Each item scores one point or zero, with agreement counting on some items and disagreement on others. Score six or more and the recommendation is referral.
Why it ended up in clinical guidance
In 2012 the UK’s National Institute for Health and Care Excellence published guideline CG142 on recognising and diagnosing autism in adults. It needed a brief case-identification tool that a GP or a mental health worker could administer without training, and the AQ-10 Test fitted that description better than anything else available.
That endorsement is why you will meet this instrument in an NHS waiting room, and why so many online screeners are copies of it. It is also the thing that researchers have since questioned.
How the AQ-10 Test is scored
Scoring the AQ-10 Test takes seconds. One point per autistic-trait response. Ten items. Maximum ten.
| Score | What the guidance says |
|---|---|
| 0 to 5 | Below the referral threshold. No further action indicated on this basis alone. |
| 6 to 10 | At or above threshold. Consider referral for a comprehensive assessment. |
That is the entire interpretive scheme. There are no bands, no profile, no severity gradient. The AQ-10 Test produces a yes or a no about whether a conversation should continue.
Which is appropriate for a triage tool. It becomes a problem when people treat the number as a verdict, because a single answer moves it by ten percent of the full range.
How accurate is the AQ-10 Test?
Here the evidence on the AQ-10 Test is genuinely poor, and it is worth laying out plainly.
In 2016 a team at the Maudsley Hospital in London examined 476 adults consecutively referred to a national autism assessment service. Seventy-three percent received a diagnosis following ADOS and ADI-R assessment. They then checked how well the short form had predicted those outcomes.
Read that specificity figure again. Twenty-nine percent. Of the people in that sample who did not have autism, fewer than one in three were correctly identified as such by the instrument.
There is a second problem, and it is structural. In 2020 a group at the University of Bath analysed data from 6,595 non-clinical adults. They found internal reliability below the accepted threshold of 0.7 on every metric they tried. A one-factor model, which is what you need if the ten items are measuring one underlying thing, fitted poorly on every index. Parallel analysis suggested four factors instead, which is unsurprising given the items were pulled from five different subscales.
Those authors drew a careful distinction. They allowed that the AQ-10 Test may still function as a clinical screening tool for diagnosable autism. They cautioned explicitly against using it as a measure of autistic traits in the general population, which is precisely what every online version of it does.
So is it useless?
No. But its value is narrower than its popularity suggests. Much narrower.
As a prompt in a ten-minute GP appointment, where the alternative is no screening at all, a two-minute questionnaire that flags three quarters of autistic people has obvious practical worth. The cost is a large number of false positives, which a subsequent assessment is supposed to filter out.
As a self-administered instrument that tells you something about yourself, it is close to worthless. Ten items cannot produce a profile. The reliability data says the items do not hang together well. And the one large study that checked it against real diagnoses found it performed at chance.
Who the AQ-10 Test is for
The AQ-10 Test was designed for adults aged 16 and over, with no intellectual disability, who are already suspected of being autistic by themselves or by a clinician. That is a narrow population. The instrument works best inside it.
There are adolescent and child versions, administered to parents rather than the young person, which are separate instruments with their own norms. Do not use the adult version for a child.
If you are taking it on your own, the honest framing is this. A score of six or more is a reason to book an appointment. A score below six is not a reason to stop wondering, and the evidence on that point is unusually clear.
Strengths and weaknesses of the AQ-10 Test
Strengths
- Genuinely fast. Two minutes, no training needed.
- Built explicitly as a referral gate and used that way in practice
- Recommended in UK national clinical guidance
- Low burden for someone nervous about starting the process
- Reasonable sensitivity, so it catches most autistic people in a referral population
- Free, short, and available in several languages
Weaknesses
- Specificity of 0.29 in the largest clinical study
- Internal reliability below 0.7 on every published metric
- Poor fit to a single-factor model, so the total score is hard to interpret
- A single answer shifts the result by ten percent
- No subscales, no profile, no developmental history
- Elevated by anxiety, which is common in people seeking assessment
Controversies around the AQ-10 Test
Whether national guidance should still recommend it
This is the live argument about the AQ-10 Test. The Maudsley authors said the recommendation may need revisiting. Others point out that no brief alternative performs demonstrably better, and that the practical choice in primary care is between an imperfect two-minute screen and nothing at all. Both positions have force.
Using a clinical triage tool as a public self-test
The Bath analysis drew a line between screening for diagnosable autism in a clinical setting and measuring traits in the general population. The instrument was validated for the first and is overwhelmingly used online for the second. That is a category error, and it is repeated on hundreds of websites.
The anxiety confound
In the Maudsley sample, generalised anxiety disorder was the only significant predictor of a higher score. If anxiety raises your score independently of autism, and anxiety is extremely common among people who suspect they are autistic, a meaningful share of positive screens are detecting distress rather than autism.
Ten items and the masking problem
Every criticism of self-report instruments for people who camouflage applies here, amplified. The AQ-50 Test at least has fifty chances to catch something. Ten items give a practised masker ten easy questions to answer in the way they have answered them their whole life.
How the AQ-10 Test compares to other instruments
| Instrument | Items | Time | What it is good for |
|---|---|---|---|
| AQ-10 Test | 10 | 2 min | Deciding whether to start a conversation |
| AQ-50 Test | 50 | 10 min | A trait profile with a far larger literature |
| RAADS-14 Screen | 14 | 3 min | Fast screening validated in psychiatric populations |
| RAADS-R Test | 80 | 15 min | Depth, and a lifetime rather than present-tense format |
| CAT-Q Test | 25 | 5 min | Explaining why your score here came out low |
| ADOS-2 Test | Observation | 40 to 60 min | Clinician-administered diagnostic assessment |
| ADI-R Test | 93 | 90 to 150 min | Developmental history from someone who knew you as a child |
| M-CHAT Screening Test | 20 | 5 min | Toddlers aged 16 to 30 months only |
If you have two minutes and want a nudge, the AQ-10 Test is the right instrument. If you have fifteen and want something you can actually reason about, the RAADS-R Test asks about your whole life rather than your present, which matters for a developmental condition.
For something broader than autism, our Neurodivergent Test covers seventeen conditions and separates traits from impact. For a plain-language autism screen in the middle of the range, the Free Autism Test takes about six minutes.
Frequently asked questions about the AQ-10 Test
I scored 6. Does that mean I am autistic?
No. It means you have reached the threshold at which referral is recommended. In the largest clinical study, about three quarters of people scoring at or above the cutoff were autistic, but that sample was already a specialist referral population where most people were. In the general public the proportion would be far lower.
I scored 4 or 5. Can I stop worrying?
Not on this basis. Sixty-four percent of the autistic people in the Maudsley study scored below the cutoff. A low score on a ten-item instrument is weak evidence, and weaker still if you mask. Take something longer before you draw a conclusion.
My GP gave me this and said I do not need a referral. Is that right?
The guidance treats a score below six as “no further action indicated on this basis alone,” and that last phrase carries weight. The screen is one input. If you have concrete difficulties affecting your work, study or relationships, say so directly and ask for them to be recorded. A questionnaire is not supposed to override a clinical history.
What is the difference between the AQ-10 and the AQ-50?
The short form is ten of the fifty items, chosen for discriminating power, two from each subscale. The AQ-50 Test gives you subscale scores and a more stable total. The short version gives you a single binary decision. Neither asks about childhood.
Is the AQ-10 Test accurate?
For what it was built to do, acceptably. For what people use it for, no. Sensitivity around 0.77 means it catches most autistic people in a referral population. Specificity of 0.29 means it flags a great many people who are not autistic. Reliability sits below the conventional threshold.
Can I use my result to get an assessment?
It can open the conversation. It will not carry it. What moves a referral along is specific examples of difficulty in specific situations, and ideally something about your childhood. A screening score alone rarely does much.
Does the AQ-10 Test work for women?
Less well. Research on sex bias in the short form has produced mixed results, but the underlying issue is well established: autistic women tend to score lower on self-report instruments while being no less autistic. With ten items there is very little room for that to average out.
Can I take it for my child?
Not this version. There are separate adolescent and child forms, completed by a parent, with their own items and their own norms. For toddlers the relevant instrument is the M-CHAT Screening Test.
Why do different websites give me different scores?
Some have transcribed the items incorrectly, some have changed the wording, and some reverse-score the wrong items. If two sites disagree by more than a point, at least one of them is wrong. The official version is hosted by the Autism Research Centre.
Is the AQ-10 Test free?
Yes, for personal and clinical use. It is reproduced in the NICE guideline, which is publicly available. Commercial use requires permission from the Autism Research Centre.
I have anxiety. Will that inflate my result?
Quite possibly. Generalised anxiety disorder was the only significant predictor of higher scores in the Maudsley sample. That does not mean your result is meaningless. It means the screen cannot tell you which of the two it is detecting, and a clinician will need to.
What should I take instead?
If you want something short but better validated for psychiatric populations, the RAADS-14 Screen. If you want depth, the RAADS-R Test. If you suspect you mask, the CAT-Q Test will tell you more than any autism screener will.
What to do with your result
Treat the AQ-10 Test as what it is. A doorbell, not a diagnosis.
If you scored at or above six, the next step is not another online questionnaire. It is writing down three or four concrete examples of difficulty, with situations attached, and taking them to a doctor. Something closer to “I have been passed over twice because I cannot do the social part of the job” than “I find socialising hard.”
If you scored below six and still feel certain, the research is on your side more than the number is. Try a longer instrument and a masking measure before you conclude anything.
And if you scored high but nothing in your life is actually suffering, that is a real and common result. Traits are not a disorder. The threshold for a clinical question is impairment, not trait count.
Sources
- Allison C, Auyeung B, Baron-Cohen S. Toward brief “Red Flags” for autism screening: the short Autism Spectrum Quotient and the short Quantitative Checklist for Autism in toddlers in 1,000 cases and 3,000 controls. Journal of the American Academy of Child and Adolescent Psychiatry, 2012;51(2):202-212. DOI · PubMed. The original development paper.
- Ashwood KL, Gillan N, Horder J, et al. Predicting the diagnosis of autism in adults using the Autism-Spectrum Quotient (AQ) questionnaire. Psychological Medicine, 2016;46(12):2595-2604. DOI · Cambridge Core. The source of every accuracy figure quoted above.
- Taylor EC, Livingston LA, Clutterbuck RA, Shah P. Psychometric concerns with the 10-item Autism-Spectrum Quotient (AQ10) as a measure of trait autism in the general population. Experimental Results, 2020;1:e3. DOI · Cambridge Core. The reliability and factor-structure analysis.
- Harrison AG, Edwards MJ. The ability of self-report methods to accurately diagnose attention deficit hyperactivity disorder: a systematic review. Journal of Attention Disorders, 2023;27(12):1343-1359. DOI. On why positive screens from brief self-report instruments are so often wrong.
Search the literature yourself
- PubMed: AQ-10 autism screening validity and brief autism screening in adults.
- Google Scholar: AQ-10 sensitivity and specificity.
- NICE CG142: Autism spectrum disorder in adults, diagnosis and management. The guideline that recommends the instrument and reproduces the items.
- Autism Research Centre, University of Cambridge. The official source for the questionnaire.
- Wikipedia: Autism-Spectrum Quotient. Background on the family of short forms.