Is therapy useful for autistic adults?

What therapy is for, and what it isn’t

The question underneath “is therapy useful” is usually “useful for what?”, and the answer isn’t autism.

Autistic adults carry a heavy load of co-occurring conditions. Croen and colleagues compared 1,507 adults with recorded autism diagnoses against 15,070 non-autistic members of Kaiser Permanente Northern California between 2008 and 2012, matched by sex and age at a 10:1 ratio. The autistic group had significantly increased rates of every major psychiatric condition examined, including depression, anxiety, bipolar disorder, obsessive-compulsive disorder, schizophrenia and suicide attempts. Read it as what it is: recorded diagnosis rates among insured adults already known to the health system, not community prevalence, and an association rather than a cause.

That study establishes why the question comes up. It doesn’t establish what therapy achieves. People also bring goals to therapy that sit outside what the studies below measured—recovering from burnout, making sense of a late diagnosis, or unpicking years of being misread. Those are reasonable things to want help with. They are not outcomes this evidence speaks to.

What the evidence does address is the co-occurring conditions. Being autistic is not a thing to be treated, and a therapist who frames it that way has misunderstood the job.

What the broader evidence shows

A 2026 preregistered meta-analysis by Smith, Holmes and White pooled randomised controlled trials and quasi-experimental studies of interventions for autistic adults, excluding single-case and pre-post designs. It covered 35 studies and 1,631 participants in total, though each result below rests on a subset of those studies.

Read it for what it measured: interventions of many kinds, including programmes targeting social cognition and academic or transition outcomes, not psychotherapy as a single modality. Its pooled figures are not a therapy-only estimate.

Three things follow from that table.

The mental health estimate is the most precise of the three, with a narrow confidence interval, and little statistical heterogeneity was detected for it, at an I² of 18.13%. It is also the smallest.

The quality of life estimate is positive, but its magnitude and generalisability are highly uncertain. Its interval runs from 0.14 to 1.15 and heterogeneity was high, at 80.48%.

The synthesis did not establish pooled improvements in behavioural, cognitive, adaptive or employment outcomes. That is not proof that no intervention helps in those areas, and those categories should not be read as measures of whether therapy makes someone less autistic.

One structural limit applies to all of it. Comparators were largely treatment as usual or waitlist, so the pooled estimates do not show that an intervention beats a specific active alternative.

What’s known about CBT specifically

CBT is the approach most often offered and the one with the most adult trial data, so it’s worth separating from the pooled picture above.

Perez-Castilla and colleagues published a 2026 systematic review and meta-analysis in Autism in Adulthood covering randomised controlled trials of CBT in autistic adults aged 16 and over, graded with GRADE. They included 10 RCTs with 537 participants, and reported that most were small, at high risk of bias, and that most of the meta-analyses showed heterogeneity.

Their conclusion is the headline: CBT showed very uncertain effects across most outcomes in autistic adults. In more detail, at under six months after treatment, CBT may reduce depression and anxiety and improve quality of life, and may have little to no effect on social anxiety, functioning, obsessive-compulsive symptoms, mental health symptom severity and all-cause discontinuation. All of those sit at very low certainty. For autistic traits, CBT may lead to an increase in scores. The one moderate-certainty finding was for alexithymia, from a single trial, where CBT probably produced an important reduction of 5.30 points, though the confidence interval ran from −10.87 to +0.27.

Two individual trials show the same picture at close range. Russell and colleagues randomised 46 adolescents and adults with autism and co-occurring OCD, mean age 26.9, to CBT for OCD or to anxiety management matched for session count. Clinician-rated OCD scores declined in both groups. The randomised comparison found no statistically significant difference between them, although responder proportions were 45% and 20%. Self-rated improvement was far smaller than clinician-rated, at effect sizes of 0.33 for CBT and −0.05 for anxiety management.

In an uncontrolled study of 19 autistic adults, Flygare and colleagues detected large clinician-rated reductions on the Yale-Brown scale, d = 1.5 after treatment and 1.2 at follow-up, but did not detect statistically significant changes in general functioning or quality of life. An uncontrolled before-and-after change is not a randomised estimate of effect, and a non-significant result is not proof that nothing changed.

What guidance actually recommends

Clinical guidance is more useful here than the trial literature, because it answers a practical question the trials don’t.

NICE, which sets guidance for England and Wales, says that for autistic adults with coexisting mental disorders, clinicians should offer psychosocial interventions informed by existing NICE guidance for the specific disorder. Staff delivering them should understand the core features of autism and their possible impact on treating the coexisting disorder, and should consider seeking advice from a specialist autism team about delivering and adapting the intervention.

On adaptations, NICE is specific. It recommends a more concrete and structured approach with greater use of written and visual information, placing greater emphasis on changing behaviour rather than cognitions, making rules explicit and explaining their context, using plain English while avoiding excessive metaphor, ambiguity and hypothetical situations, involving a family member, partner, carer or professional with the person’s agreement, and maintaining attention through regular breaks and by incorporating special interests.

The honest limit on all of that: these adaptations are recommended and widely described, but studies rarely isolate them as the thing being tested. Deakin and colleagues found that few of the studies in their review identified how adaptations affected wellbeing outcomes. Moore and colleagues, synthesising 13 studies of mental health professionals’ experiences, found clinicians treat adaptation as a highly individualised process shaped by their own experience and by what their service permits. So adaptation can make therapy more accessible and more acceptable; how much it changes clinical outcomes is something few of these studies isolated.

How thin the research still is

Two reviews put numbers on it.

Deakin and colleagues screened 4,186 records and included 69 papers: 42 adult-only studies and 27 with mixed adult and younger samples. Forty-two of the 69 were pilot or feasibility studies. Research into psychological wellbeing interventions for autistic adults, in their framing, is still principally assessing feasibility.

Timmerman and colleagues looked at what those trials measure. Screening over 10,000 records, they found 19 randomised trials of mental health interventions for autistic adults, and reported that only 5 measured quality of life as an outcome. Autistic adults have said quality of life is the outcome that matters most to them in mental health research.

Finding someone, in practice

Access is its own barrier, and it has been measured. Nicolaidis and colleagues surveyed 437 people in a US community-based participatory study, of whom 209 were autistic. Among the autistic participants, the most-selected barriers were fear or anxiety, at 35%, not being able to process information fast enough to take part in real-time discussion, at 32%, concern about cost and sensory issues caused by facilities, both at 30%, and difficulty communicating with providers, at 29%. That was a self-selected community sample rather than a representative one, so read the ranking rather than the exact rates.

Several of those barriers can be discussed before a first appointment. Ask whether you can book by email rather than phone. Ask whether questions can be sent ahead in writing. Ask what the waiting room is like and whether you can wait elsewhere. Ask how the therapist handles communication, pacing, written information and sensory needs, what experience they have with autistic adults, and whether they seek specialist advice when a case calls for it. The answers give you useful information that a credential alone may not.

On cost, some of the programmes described on our page about what to do if you can’t afford an assessment may also fund mental health services, but eligibility, covered services and local availability differ, so confirm therapy coverage with the programme itself. If you’re weighing whether a therapist can also confirm a diagnosis, that’s a separate question: can a therapist diagnose ADHD or autism.

What the research can tell you is narrow and still worth knowing. Therapy has limited and uncertain evidence for the conditions that co-occur with being autistic, and none for changing autism itself. Decisions about treatment, safety, medication or changing current care belong with a qualified clinician who knows your circumstances.