Is Video Game Addiction Real?

The honest answer has two halves, and most coverage only carries one of them.

Yes, in the narrow sense: the World Health Organization recognises a formal diagnosis, a small number of people meet it, and they are genuinely unwell. No, in the broad sense the phrase usually implies: the best causal evidence finds that how many hours someone plays does not predict how they feel, the prevalence figures are largely an artefact of which questionnaire gets handed out, and the leading scientific objection to the whole category has never been answered.

Both halves matter if you are trying to work out whether someone you know has a problem.

What the WHO actually classifies

Gaming disorder entered the eleventh revision of the International Classification of Diseases as code 6C51. The WHO's own description gives three features:

1. Impaired control over gaming.
2. Increasing priority given to gaming over other activities, to the point that gaming takes precedence over other interests and daily activities.
3. Continuation or escalation of gaming despite negative consequences.

Then comes the clause that does most of the work, and that almost never survives into a headline. The pattern must be severe enough to cause significant impairment in personal, family, social, educational or occupational functioning, and would normally have been evident for at least 12 months.

The WHO adds, in the same document, that studies suggest gaming disorder affects only a small proportion of people who play.

There is a second code most people have never heard of, and it changes how this should be read. Hazardous gaming, QE22, describes a pattern that raises the risk of harm and warrants advice from a health professional, but explicitly does not meet the requirements for gaming disorder. It sits in the chapter on reasons for contact with health services rather than among the mental disorders. It is a risk state, not an illness.

Almost every worried-parent conversation, and almost every prevalence survey, collapses those two tiers into one. That collapse is the source of most of the confusion below.

The American Psychiatric Association has not agreed

This is the part that makes "officially recognised as a mental illness" only half true.

The APA's own patient guidance states that internet gaming disorder appears in the section of the diagnostic manual that recommends conditions for further research, alongside caffeine use disorder. It is not a formal diagnosis. The proposed criteria list nine symptoms, of which five within a year would be required, and the APA notes the criteria cover gaming only, not general internet use, online gambling or social media.

The two bodies do not merely disagree about whether to include it. They disagree about what it is. The APA's draft criteria include tolerance and withdrawal, borrowed from substance dependence. The WHO deliberately left those out and anchored its definition to loss of control and functional impairment instead.

The APA also states that researchers have estimated 0.3 to 1.0 per cent of the general population might qualify for a potential diagnosis. Worth holding next to the numbers in the next section but one.

Serious scientists argued hard against the whole category

In 2017, twenty-five researchers published an open debate paper in the Journal of Behavioral Addictions asking the WHO to drop the proposal. Their argument, in their words: the quality of the research base is low, there is no consensus on symptoms or assessment, the criteria lean too heavily on substance and gambling models, and a current moral panic about video games risks pushing the medical community into ill-considered steps. Criteria with low specificity, they warned, would misclassify large numbers of ordinary players.

A follow-up the next year, signed by thirty-six researchers, put the sharpest version of the objection. Is what we are calling gaming disorder actually a coping strategy for depression, ADHD or anxiety? If gaming is a coping behaviour, they argued, the sensible move is to investigate the underlying cause first. They also identified a group they called highly engaged players, somewhere between 1.1 and 10.9 per cent of the samples they reviewed, who endorse addiction-scale symptoms while showing no measurable impairment in their lives.

The other side is not a straw man and deserves its own paragraph. Researchers arguing for inclusion pointed out that people are already seeking treatment and suffering real functional impairment, that specialist services in Germany quadrupled between 2008 and 2015, and that a formal code is what enables clinician training, funding, monitoring and research. Their bluntest point: the harm from excluding a condition that demonstrably burdens people is greater than the harm from including it. Others noted that the ICD-11 wording deliberately kept only the least contested criteria, and that a diagnosis can reduce stigma by reframing a problem as a condition rather than a character flaw.

Nobody in this argument denies that a minority of people are badly impaired. The dispute is whether that impairment is best modelled as an addiction, whether the questionnaires can tell an impaired player from an enthusiastic one, and whether formalising the label does more good than harm.

Why the prevalence figures are useless as stated

If you have seen a confident percentage for how many gamers are addicted, it was unearned. Here is why.

A meta-analysis of 53 studies covering 226,247 people put pooled prevalence at 3.05 per cent. Its important finding was not that number. It was that the choice of screening tool alone explained 77 per cent of the variation between studies. The authors concluded that prevalence rates appear inflated by methodological characteristics, particularly measurement and sampling.

A 2024 meta-analysis demonstrated the point directly by sorting studies by which questionnaire they used.

Sorted by the questionnaire used, reported prevalence runs 2.7 per cent on the IGDT-10, 3.0 on the GADIS-A, 6.6 on the IGDS9-SF, 10.2 on the IGD-20 and 13.7 on the DSM-5 checklist. The same construct, in broadly similar populations, comes out at 2.7 per cent or 13.7 per cent depending on which form you hand out. Split by criteria rather than instrument, studies using DSM-5-style criteria produced 7.9 per cent and studies using ICD-11 criteria produced 3.0 per cent. Adjusting the overall figure for publication bias roughly halved it, from 6.7 to 3.3 per cent, which is its own warning about the literature.

The high-end figures are worse than imprecise. A 2024 meta-analysis of 84 adolescent studies covering 641,763 young people reported a pooled 8.6 per cent, with a prediction interval running from 1.0 to 48.1 per cent and heterogeneity effectively total. One of its component estimates put 61.3 per cent of Egyptian adolescents in the category, which is not a finding about Egypt. It is a measuring instrument breaking.

The defensible summary: roughly 1 to 3 per cent under conservative, impairment-anchored measures, considerably more under loose symptom checklists, and the difference is about the questionnaire rather than the players.

What happens when you measure actual play time

Almost all of the research above rests on people describing their own gaming. A smaller body of work does not, and it points somewhere else.

Researchers at the Oxford Internet Institute obtained objective play-time telemetry from seven publishers, covering 38,935 players across games including Animal Crossing, Gran Turismo Sport, Forza Horizon 4, Apex Legends and Eve Online, and matched it to three waves of well-being surveys over six weeks. They used a model that separates change within a person from stable differences between people, which is what kills most screen-time correlations.

Play time did not predict well-being in either direction. The estimates sat on top of zero. To produce the smallest difference a person would actually notice on their affect measure, the authors calculated, someone would need to play about ten hours a day more than they usually do.

What did predict well-being was why people played. Intrinsic motivation, playing because the game is absorbing, was reliably linked to higher life satisfaction. Playing out of compulsion or obligation pointed the other way. Motivation mattered; duration did not.

There is a quasi-experiment too. During the pandemic, Japanese retailers allocated scarce consoles by lottery, which is as close to random assignment as this field gets. A study of 97,602 people, 8,192 of whom had entered those lotteries, found that winning a console modestly reduced psychological distress, by 0.08 to 0.18 standard deviations depending on the console. Gaming came out slightly good for people, not bad.

Two cautions on that one. It measures the effect of acquiring a games console during a lockdown, which bundles novelty and social contact with the gaming itself. And the benefits diminished once play exceeded about three hours a day, which is a ceiling rather than a licence.

What does reliably track badly is sleep

The strongest practical case for limits is not addiction. It is displacement, and sleep is where it shows up.

A survey of 3,228 Norwegian adolescents found that those classified as addicted gamers slept about 34 minutes less on weekdays than other teenagers, and that problem gamers had roughly twice the odds of difficulty falling asleep. A detail worth noticing: the highly engaged but unimpaired group showed as much disruption to sleep timing as the addicted group, without the sleep-duration deficit. The signal tracks when people play more than whether they are ill.

It is also cross-sectional, so it cannot say whether late gaming causes poor sleep or poor sleepers game late. Both are plausible and they probably feed each other.

What a clinical case actually looks like

England has a specialist NHS service for this, and it published a profile of its first 380 gamer referrals. The average person referred was 19 years old, 60 per cent were aged 13 to 18, 90 per cent were male, and they were gaming an average of ten hours a day. One in ten had a formally diagnosed neurodevelopmental condition and one in eight had an existing mental health condition.

That is the shape of the real thing. Ten hours a day, a family in crisis, functioning collapsed, and in a substantial minority of cases something else diagnosed underneath. It is a long way from a teenager who plays for three hours after homework.

The warning signs that actually matter

Take them from the diagnostic definition rather than from the clock, because the clock is exactly what the definition avoids.

• Loss of control, repeatedly, over starting and stopping rather than on one bad evening.
• Displacement to the point of precedence, where gaming has pushed out the things that used to matter rather than merely competing with them.
• Continuing or escalating despite consequences that the person themselves recognises.
• Real functional damage at school, at work, in the family or in friendships.
• Persistence, normally for around a year, not a fortnight after a game launched.

A person who plays a great deal, sleeps enough, keeps their friends, turns up to school or work and can stop for a family holiday does not meet any part of that.

Bans do not work, and there is unusually good evidence

China imposed one of the strictest gaming restrictions in the world, limiting under-18s to 90 minutes a day with a night-time curfew. Researchers then examined roughly 7.04 billion hours of play recorded in industry telemetry across mainland China in a pre-registered analysis.

Heavy play did not fall. It rose very slightly, from 0.77 to 0.88 per cent of weekly player profiles, a change far below the threshold the researchers had committed in advance to treating as meaningful. Their data could not separate minors from adults, and minors may have used adult credentials, which is itself part of the answer about what bans achieve. South Korea's overnight shutdown law for under-16s was repealed in 2021.

The American Academy of Pediatrics has moved the same way, stating there is not enough evidence behind specific screen-time limits and that rules focusing on balance, content, co-viewing and communication are associated with better outcomes than rules focused on time. Our guide to what to do when the phone rules get broken is built on the same logic, and the family phone agreement is where most households should start rather than with a ban.

Loot boxes are the part with the clearer case

If you want the genuinely worrying mechanic, it is not the games. It is the payments inside them.

A systematic review found loot box spending associated with problem gambling in twelve of thirteen studies, at a mean correlation of about 0.27, and with problem gaming at about 0.40. The authors are explicit that direction of causality is not established.

The UK government's own review of the evidence, published in July 2022, found a stable and consistent association across fifteen peer-reviewed studies, with emerging evidence that greater spending tracks greater problem gambling severity, while stating plainly that causation has not been established. It declined to legislate at that point in favour of industry-led protections. The figure from it worth remembering: the top 5 per cent of spenders generated half of all loot box expenditure, which is the same revenue concentration seen in gambling.

If you are worried about someone

Five things, in the order they are likely to help.

1. Look at sleep before you look at hours. It is the displacement with the clearest evidence, the easiest to observe and the most likely to be fixable on its own.
2. Ask what has been given up, not how long they played. The diagnostic standard is precedence and impairment. A shrinking life is the signal; a large number on a clock is not.
3. Treat a sudden escalation as a symptom rather than a cause. The strongest scientific objection to this whole category is that much problem gaming is a coping response to depression, anxiety or an attention difficulty. If gaming climbed after something else changed, look at the something else. Our summary of what the evidence says about teenage mental health and screens covers how weak the direct link turns out to be.
4. Negotiate structure rather than prohibition. The largest test of a hard ban found it changed nothing measurable, and the paediatric guidance favours rules about balance and content over rules about minutes. If controls are part of your plan, our look at whether parental control apps actually reduce problem phone use is worth reading first.
5. Know that treatment evidence is thin but not empty. A meta-analysis of psychological treatments found large effects that shrank substantially when compared against an active control rather than no treatment, from a Hedges' g of 1.70 down to 0.88, on a base of only seven randomised trials. The single most credible study is a Swedish randomised trial of 102 adolescents in real psychiatric clinics, where a short course of relapse-prevention therapy beat treatment as usual with a moderate to large effect. Promising, small, and not yet a settled treatment.

The line worth defending

Video game addiction is real in the sense that a small minority of people meet a formal WHO diagnosis defined by lost control and collapsed functioning, and they get treated for it. It is not real in the sense the phrase usually carries, because hours played does not predict well-being, the prevalence numbers are mostly a property of the questionnaire, hard bans demonstrably do not work, and the objection that most problem gaming is a symptom of something else has never been refuted.

If someone in your house is gaming in a way that frightens you, none of this is a diagnosis and none of it is medical advice. Persistent functional decline, whatever the apparent cause, is worth taking to a GP or a mental health service rather than to a stricter router setting.