What the
research shows
about social
media and
self-esteem
in autism
and ADHD

Sixteen studies and a split that follows the neurotype

Anna O’Leary, Alice Potter, Kieran Rose, Georgia Pavlopoulou and Marc Tibber — a UCL team with lived and family experience of autism and ADHD on it, and Rose a well-known autistic advocate — have published a scoping review in Autism in Adulthood on social media use and self-esteem in autistic people and people with ADHD. Sixteen studies met the criteria: eleven on children and adolescents, five on adults; ten on ADHD, five on autism, one on both. Pre-registered, PRISMA-compliant, and — this matters for what follows — read for what kind of question each study asked as much as for what it found.

The headline findings look, at first, like a clean neurotype split. In people with ADHD, social media use was consistently linked to lower self-esteem: problematic or addictive patterns of use tracked with more ADHD traits and worse self-esteem across risk profiles in three separate studies, and cyberbullying — victimisation and perpetration alike — sat with lower self-esteem in three more. In autistic people the picture was mixed: cyberbullying was again associated with lower self-esteem, but online connection and community were associated with higher self-esteem — autistic adults describing how seeing other autistic people share their lives, unashamed of their diagnosis, gave them a better self-image and a sense of not being alone.

Read at face value, that would suggest social media works differently for the two populations — corrosive for ADHD, double-edged for autism. And there may be some truth in it: the review notes that certain ADHD traits, impulsivity above all, interact with design features like the infinite scroll to drive prolonged and ultimately problematic use, and that ADHD adults describe the burden of continuous written communication and the fear of rejection that comes with it. But the authors then do something the field rarely does. They look at what the sixteen studies were built to see — and the split largely dissolves.

Or a split that follows the question the researchers chose to ask

Here is the pattern. Every one of the ten ADHD studies was quantitative. Seven of them looked at problematic or addictive use; three looked at cyberbullying. Not one asked about community, connection, identity, or benefit. So the ADHD literature could only ever return findings about harm, because harm was the only thing it measured. The autism studies, by contrast, were more likely to be qualitative, more likely to ask about general use and online experience — and only there could benefit surface, because only there was there room in the instrument for it. The two studies that found a positive relationship between social media use and self-esteem were the two studies of diagnosed autistic adults, asked open questions.

The authors say it plainly: it is unclear whether the relative absence of identity-affirming and community-based mechanisms in the ADHD literature reflects genuine differences in lived experience or limitations in research framing. Most ADHD studies were structured so that “it was not possible for potential benefits to emerge.” And they go further, naming the wider disposition of the field — a “concern-centric” focus, “causationist” assumptions, catastrophising narratives that conceptualise social media as implicitly harmful and operationalise it reductively: time spent, number of platforms, addiction scores, presence or absence of bullying. Very few studies measured motivation, context, meaning, or the function that heavier use might be serving.

This is the argument from the van Asselt piece — that a coding scheme can encode its verdict before it meets the data — arriving as an empirical twin. The instrument decides what can be found. Measure only addiction and bullying, and you will document addiction and bullying, and you will conclude that social media lowers self-esteem in this population; and you will be right about what you measured and wrong about what you concluded, because you never asked whether the same platforms were also where an ADHD teenager first found someone describing their own mind. The autism literature was permitted to find community because a handful of researchers asked about it. The ADHD literature was never permitted to. And the review’s sharpest observation is that this isn’t neutral: framing online harm as the consequence of individual vulnerability — the impulsive ADHD user, the socially naïve autistic one — rather than as the product of platform design, algorithmic amplification, intersectional stigma, and the scarcity of safe online spaces, is the deficit model migrating from the clinic to the feed.

Holding both — the feed that captures and the community that mirrors

This review requires me to hold two of my own arguments together, and I would rather do it in the open. Three weeks ago, in the piece on pathological demand avoidance, I argued that the frictionless screen-world deepens demand-avoidance and forecloses the return to the real one — that the feed is a libido sink, spending the desire it appears to restore. I stand by that, and this review supports it precisely where it applies: problematic, compulsive, infinite-scroll use, driven by impulsivity, tracks with lower self-esteem, and the mechanism the authors describe — impulsivity meeting design engineered to exploit it — is the mechanism I described.

But the same review documents the opposite effect from the same class of technology, and it is not a small effect: autistic adults’ self-esteem rising from online community — from seeing others unashamed, from finding people who share their way of being, from what the authors call the value of an identity-based explanatory framework and access to a community. And, in the same qualitative study, self-esteem falling when the autism content encountered was stereotyped or medicalised. The variable was never the platform. It was what came through it: capture or mirror, algorithm or community, a stereotype of you or a person like you.

So the honest position is the discrimination again — the same one the van Asselt piece asked for. Not “is social media good or bad for neurodivergent people,” which is unanswerable because it isn’t one thing, but which social media, doing what, to whom. The feed that captures attention it did not earn is corrosive, and the corpus has said so. The community that reflects a self back unashamed is restorative, and this review shows it. Both run on the same devices, sometimes in the same app, sometimes in the same hour. A research programme that can only see one of them will keep publishing half the truth as if it were the whole — and a parent, or a clinician, or a commentator working from that literature will keep prescribing abstinence where the right prescription might have been a better community.

What would have to be measured before anyone could actually know

The review is candid about the state of the evidence, and the candour is the point. Thirteen of sixteen studies were cross-sectional, so nothing here establishes cause or direction — low self-esteem may drive problematic use as readily as the reverse. Fourteen relied on self-report of technology use, which correlates poorly with device data. Most samples were adolescents; adults, who may use social media differently and have more stable self-esteem, were thin. Eight of the ten ADHD studies used general-population samples measuring traits, not diagnosed people — which matters, because the two studies that found benefit were the two of diagnosed autistic adults, suggesting that having a name for your difference and a community organised around it may be exactly what converts a platform from threat to resource. No study of diagnosed participants included a neurotypical control group. Ethnicity went unreported in more than half. Gaming was excluded by design. And co-occurrence was almost never recorded, so the “autism” and “ADHD” columns quietly contain an unknown number of AuDHD people sorted by which label the researcher was interested in at the time.

The agenda the authors set follows from all of that: longitudinal and experimental designs that can speak to causality; qualitative work that captures function and meaning; richer constructs than addiction and bullying; diagnosed samples with controls; adults; and — in line with autistic and ADHD community priorities from their own earlier co-produced Delphi study — deliberate oversampling of minoritised and intersectional identities, who stand to gain most from online community precisely because they are most excluded offline.

What the review finally shows is less a fact about social media than a fact about how a field decides what counts. Sixteen studies, and the finding depended less on the neurotype than on the shape of the question. Ask about harm and you find harm; ask autistic adults what online life is like and they tell you about the first time they saw someone unashamed of the thing they’d been taught to hide. The task now — for researchers, and for the rest of us reading them — is to build instruments wide enough to hold both answers, because the people using these platforms have been living both at once the whole time.

Citations

O’Leary, A., Potter, A. L., Rose, K., Pavlopoulou, G. & Tibber, M. S. (2026) — The relationship between social media use and self-esteem in Attention Deficit Hyperactivity Disorder and autism — a scoping review — Autism in Adulthood (accepted manuscript; protocol: OSF)

Tibber, M. S., Rose, K., Mandy, W., Pellicano, E. & Pavlopoulou, G. (2025) — Autistic adults’ priorities for social media and mental health research: a coproduced Delphi consensus study — Autism in Adulthood

Skafle, I., Gabarron, E. & Nordahl-Hansen, A. (2024) — Social media shaping autism perception and identity — Autism

Orben, A. (2020) — The Sisyphean cycle of technology panics — Perspectives on Psychological Science

Triantafyllopoulou, P., Clark-Hughes, C. & Langdon, P. E. (2022) — Social media and cyber-bullying in autistic adults — Journal of Autism and Developmental Disorders