The social media anxiety study 2025 conversation has acquired a second front this year, not about whether platforms cause harm, but about whether the label itself has stretched too far. Social media anxiety is the cluster of worry, comparison and physiological arousal that builds in heavy or compulsive users of platforms like Instagram, TikTok and Snapchat, but it is not a formal diagnosis in any clinical manual. That linguistic gap is now the centre of a real academic debate. Two careful research camps disagree, in print, about whether ordinary teen distress is being relabelled as disorder, and whether the relabelling itself is doing harm. This piece walks through what each camp argues, where they actually agree, and what a defensible middle position looks like for anyone reading the headlines.
The argument is not a flame war. It is a respectful disagreement between credentialled researchers with shared concern for adolescents, and it is exactly the kind of debate the field needs to mature past its first decade of cross-sectional surveys. Our anxiety topic hub sets the wider clinical context if you want to ground the conversation in standard categories before reading on. The shorter version of where this article lands: anxiety disorders are real and well-defined, social media is a meaningful environmental risk factor for a meaningful subgroup, and the casual application of clinical labels to ordinary teen distress almost certainly outpaces what the 2025 evidence justifies.
What overdiagnosis means here
Overdiagnosis, in clinical medicine, is the application of a disease label to people whose findings would never have caused them harm if left alone. In a mental health context the line is fuzzier, but the underlying worry is similar: applying disorder language to people whose distress is real but transient, normal, and unlikely to meet formal criteria on closer inspection. For social media anxiety, the concern from contrarian researchers is that a vivid, intuitive label has begun to attach to almost any negative feeling associated with scrolling, and that this drift is happening faster than the supporting evidence.
To be precise about what is and is not being claimed: nobody serious in this debate denies that anxiety disorders exist, that they can be exacerbated by social media use, or that a meaningful subgroup of teens are genuinely struggling. The disagreement is about scale, threshold and labelling. It is also about which interventions follow, clinical or behavioural, individual or platform-level, treatment or design reform.
A second clarification matters here. “Social media anxiety” is not a recognised diagnostic category in either the DSM-5-TR or the ICD-11. The closest formal labels are generalised anxiety disorder, social anxiety disorder, and various stress and adjustment categories, all of which require specified symptom thresholds, durations and functional impairment to be met. When journalists, parents, and increasingly teens themselves talk about “having social media anxiety”, they are using a popular phrase that may or may not correspond to any of those formal categories. The looseness of the term is part of what makes the overdiagnosis worry concrete: an informal label can spread far more easily than a manualised one because nobody is checking the criteria.
The case that overdiagnosis is happening
The Fassi, Orben, Przybylski, Ford and Ferguson 2025 paper in Nature Human Behaviour is the most cited single piece of evidence for the overdiagnosis worry. The team analysed nationally representative UK data on adolescents with and without diagnosed mental health conditions, and found two things that, taken together, complicate the universal-harm story. First, the average effect of social media use on mental health symptoms across all teens is small. Second, the effect is concentrated in a subgroup of teens who already have a diagnosis, and even there, the direction of causation is partially reversed, with mental-health-affected teens drawn to more intense and emotionally reactive use.
Their inference is not that platforms are harmless. It is that population-level harm claims, of the kind that drive media panic and parent worry, outrun what the population-level data show. If only a subgroup is meaningfully affected, then applying a “social media anxiety” label across the average teen is, by definition, overdiagnosis. The Orben group’s preferred framing is that platforms are an environmental risk factor that interacts with prior vulnerability, not a free-standing pathogen that hands out a new disorder to everyone with a phone.
The 2024 digital exposome paper in PMC11504934 fits this case. It found that the relationship between social media use and youth mental health depends heavily on patterns of use, content type, time of day, social context, rather than on a single global exposure metric. That granularity argues against a single broad label and for finer-grained behavioural targets. Treating “social media anxiety” as one thing collapses a distinction the exposome work shows is doing real explanatory work.
A further strand of the overdiagnosis argument is sociological rather than statistical. Anxiety language has become the lingua franca of adolescent self-description over the past decade. Words that used to belong inside a clinic, “trigger”, “trauma”, “panic attack”, “anxiety spiral”, now appear in casual conversation, on social platforms themselves, and in school counsellor intake forms. That diffusion has a humane upside: it has reduced stigma and made it easier for genuinely struggling teens to ask for help. But it also creates the conditions in which any unpleasant experience around a phone can be slotted into clinical vocabulary, even when the underlying physiology and impairment do not meet the threshold the vocabulary was built for. The overdiagnosis camp is asking the field to take that semantic drift seriously when reading the prevalence statistics.
The case against, Sigaud, Haidt and the 2026 commentary
The most thorough counter-argument arrived in 2026 with the Clinical Psychological Science commentary by Sigaud, Rausch, McClean and Haidt. The team critiqued three frequently cited studies by Vuorre and Przybylski, work that has been read as supporting the “small effects” position, and argued that methodological choices in those papers systematically understate the real impact of social media on adolescent mental health. Their list of concerns includes how exposure is measured, how outcomes are operationalised, how cohort effects are modelled, and how the null findings are framed for public consumption.
The implication for the overdiagnosis debate is significant. If the “small effects” headline that supports the overdiagnosis worry rests on studies with measurement and modelling problems, then the average effect may not be as small as critics of alarm claim. In that reading, the diagnostic label is not being applied too widely; it is, if anything, still under-applied relative to a harm signal that careful measurement would surface. Sigaud and Haidt’s broader public position has been that the harms to adolescent mental health since the platform shift are real, structural and underestimated, and that the methodological caution championed by the Orben group has been weaponised to slow policy responses.
It is worth saying clearly: the Sigaud/Haidt camp is not arguing for reckless labelling. They are arguing that population-level effect sizes are larger than the small-effects critics report, and that clinical and policy responses should scale accordingly. Whether that response should include broader diagnostic application of anxiety labels is a separate question they handle with more care than some of their popular followers.
A useful test of who has the better of this exchange is what each camp predicts will happen as measurement improves. The Orben camp expects effect sizes to stay small or shrink as cohort designs, pre-registration and better exposure measurement reduce noise. The Sigaud and Haidt camp expects effect sizes to grow as measurement catches up to the real impact. The next two to three years of literature will distinguish those predictions empirically. Until then, prudent reading is to hold both possibilities open and to weight neither label-spreading nor label-restricting positions with more confidence than the data warrant.
Where the two camps actually agree
For all the heat in print, the agreements between the Orben and Haidt camps are larger than the disagreements. Both agree that adolescent mental health is in a worse place than it was a decade ago. Both agree that social media is a meaningful environmental risk factor. Both agree that some subgroups, teen girls, teens with prior mental health conditions, teens with heavy or problematic use patterns, are at materially higher risk. Both agree that platform design choices are doing work that nobody voted on. And both agree that careful measurement matters.
The narrow disagreement is about three things: how large the average population-level effect is, how much of that effect reflects causation versus selection, and whether the right policy response is universal (platform reform, age limits, content rules) or targeted (clinical care for the affected subgroup, behaviour change for the at-risk subgroup). That is a real disagreement with real consequences, but it is not the screaming match that headline coverage sometimes implies. Our contrarian read of the panic narrative covers the same disagreement from a different angle, about how the panic itself gets translated into headlines, and is a useful companion read.
What “social media anxiety” should and shouldn’t cover
A defensible working definition for this site, and for parents and clinicians trying to use the term carefully, is that “social media anxiety” describes a behavioural pattern of worry, comparison and arousal around heavy or compulsive platform use that is intense enough to interfere with sleep, school or relationships. That definition is narrower than the casual usage now common in social media discourse, but broader than any formal diagnostic category.
By that working definition, the label legitimately covers a meaningful minority of heavy users, particularly teen girls and teens with prior anxiety vulnerability. It does not legitimately cover the much larger group of teens who occasionally feel bad after looking at Instagram and then move on with their day. That second group is having a normal emotional experience of growing up online, not a disorder, not a behavioural addiction, and not, on any defensible reading, social media anxiety.
The distinction is not academic. A teenager who feels a pang of envy at a friend’s holiday photos and is fine two minutes later is not a clinical case. A teenager who cannot put the phone down at three in the morning, whose grades have slipped, whose friendships have narrowed, and who feels physically panicked when separated from notifications is a clinical case, though even there the right label may be a recognised anxiety disorder rather than a free-standing “social media” diagnosis. The Sigaud and Haidt commentary would argue that the second teenager is more common than the small-effects critics acknowledge; the Fassi and Orben work would argue that the first teenager is more common than the alarmists imply. Both can be true, and the right calibration probably depends on which population you are looking at.
This framing is also what the 2024 digital exposome data support. The strongest signals there sit with specific patterns of use, not with the existence of any negative feelings. The label, used carefully, names a pattern that needs intervention. Used loosely, it names a normal experience that does not.
The harms of getting the label wrong
If the contrarian researchers are right that overdiagnosis is happening, the costs are not abstract. Misallocated treatment capacity is one, clinicians have limited hours, and routing the worried well into anxiety care leaves less time for severe cases. Identity formation is another: teens taught to interpret normal discomfort as pathology often integrate the diagnosis into how they see themselves, which can entrench an illness identity that does its own anxiogenic work. Population-level meaning is a third: when “anxiety” applies to most of the population, the word stops differentiating between mild discomfort and the genuinely disabling condition that the diagnostic manual was written for.
There is one further harm worth naming. Medicalising the social media experience can quietly let platform design off the hook. If the framing is “your daughter has social media anxiety”, the implied solution is treatment for her. If the framing is “platforms are engineered to produce a behavioural pattern that drives anxiety in many users”, the implied solution is platform reform. The first framing puts the burden on the affected individual; the second puts it on the system. Diagnostic inflation, even when well intentioned, can shift that burden in unhelpful ways. Our overview of the wider evidence FAQ covers the standard questions parents ask when they are trying to decide where to put the burden in their own family.
A defensible middle position
Reading both camps closely, the position the evidence will tolerate is roughly this. Anxiety disorders are real, measurable and well-defined; the threshold for diagnosing them should remain tight. Social media is an environmental risk factor for anxiety symptoms in a meaningful subgroup, particularly teen girls with prior vulnerability and any user with a problematic-use pattern. Treating the platforms as one important risk factor among several, on the same conceptual footing as sleep loss or alcohol use, is empirically supported and clinically useful. Treating exposure as a disorder in its own right is empirically thin and clinically costly.
That middle position keeps the diagnostic system honest while taking the platforms seriously. It also matches the way parents and clinicians actually function best, naming the behaviour, changing the behaviour, watching for genuine disorder, and treating the disorder if it appears. Our mothers-and-teens piece covers the surprising data on which generation is more affected, and is a good read for parents trying to apply this calibration at home.
Frequently asked questions
The FAQ above this section is the structured-data version Google reads. The body section ends here. If you would like to keep reading the surrounding debate, the most natural next stops are our panic-vs-evidence contrarian read, the biggest findings of 2025 ranked, and the Fassi 2025 deep dive on how anxious teens use platforms differently.