Home › Player Health

Video Game Addiction Statistics 2026: What the Numbers Actually Measure

Nearly every page on this subject quotes one addiction rate as if a rate exists. It does not. Published prevalence estimates for gaming disorder span close to two orders of magnitude, and the spread is not measurement noise — it is definitional. Change the questionnaire and the answer moves more than changing the country does. The single most useful thing to know about any video game addiction statistic is therefore not the number: it is that the number tells you which instrument was used, not how many people are ill. This page lines the instruments up against their outputs and shows exactly how much of the disagreement they explain.

What does almost every article on this topic get wrong?

It treats “problematic gaming” and “gaming disorder” as the same measurement, when one is a symptom count and the other requires documented harm lasting about a year. That single conflation generates most of the confusion in this field.

The World Health Organization put gaming disorder into ICD-11 with three behavioural criteria — impaired control, escalating priority over other activities, continuation despite negative consequences — plus a fourth, separate requirement: the pattern must cause significant impairment in personal, family, social, educational or occupational functioning, and must normally have been evident for at least 12 months. The American Psychiatric Association took a different route: Internet Gaming Disorder sits in DSM-5 Section III, the appendix for conditions needing further study, defined by meeting five of nine proposed criteria.

Most published prevalence numbers come from neither. They come from self-report questionnaires administered once, scoring symptoms that a person may well answer “yes” to while holding down a job, a degree and a relationship. WHO’s own guidance is blunt about the scale of the real problem: studies suggest gaming disorder affects only a small proportion of people who play. The literature that produces the alarming headlines is largely not measuring that group.

Myth 1: “About 9% of gamers are addicted”

Reality: 8.6% is a pooled adolescent screening figure with 100% heterogeneity across its 84 constituent studies — it is a description of a literature, not of a population. Line up the major estimates and the pattern is immediate.

Prevalence definitions differ by row: some are general-population rates, some are rates among players only, and the age bands differ. Mixing them is one of the errors this page is about, and the rows should not be averaged. Croatian fieldwork was conducted in 2019 and published in 2025. Stevens et al. pooled 226,247 participants across 53 studies in 17 countries; a 2023 corrigendum removed two studies that had used percentile rather than 5-of-9 scoring, revising the pooled figure slightly downward. Sources listed in full below.

The Borges rows are the ones to stare at. Same 7,022 people. Same questionnaire session. Two scoring rules. 5.2% under DSM-5, 2.7% under ICD-11 — and the authors note that adding an impairment requirement shrinks both, while DSM-5 stays the larger of the two. No epidemiological change can explain a gap produced by choosing which manual to score against.

Myth 2: “The estimates disagree because the world is changing”

Reality: the largest global meta-analysis quantified it — the choice of screening tool alone accounted for 77% of the variance between studies. Stevens and colleagues, publishing in the Australian and New Zealand Journal of Psychiatry in 2021, pooled the world’s prevalence literature at 3.05%. Restrict the pool to studies using rigorous sampling — stratified random selection rather than whoever answered an online call for gamers — and the estimate falls to 1.96%, with a confidence interval so wide (0.19% to 17.12%) that the honest reading is: the literature does not yet know.

Their moderator analysis named the culprits precisely. Instruments such as the Lemmens IGD-9, the Gaming Addiction Identification Test and the Problematic Videogame Playing scale produced the highest estimates. Adolescent samples, lower cut-off scores and smaller studies all pushed prevalence up. The 2026 young-adult meta-analysis in Addictive Behaviors found the same fingerprint from the other direction: DSM-5-based tools returned higher rates than the ten-item IGDT-10, and gamer-only samples returned 8.1% against 5.47% for mixed samples of gamers and non-gamers.

There is one genuine temporal signal worth naming honestly: the adolescent meta-analysis found a significant positive association between prevalence and year of publication. Rates in the literature are drifting up. Whether that reflects more disordered gaming, more sensitive instruments, or a research field that increasingly recruits where the problem is, the study cannot say — and neither can we.

Myth 3: “We know how many people worldwide are affected”

Reality: apply the credible player-denominated rates to the world’s 3.6 billion players and the answer lands anywhere between 59 million and 310 million people. That is not a forecast disagreement. It is the same year, the same planet, the same people — measured with different rulers.

16Best analysis. Each rate is applied to the global player base of about 3.6 billion (see our video game industry statistics). This is deliberately crude: age-specific rates should not be applied to a whole-population denominator, which is precisely why the resulting spread is the point rather than any single row.

Myth 4: “Heavy players are addicted players”

Reality: behavioural data and symptom questionnaires disagree by more than an order of magnitude about how common extreme play is. In 2023, David Zendle and colleagues published a study in Nature Human Behaviour built on telemetry rather than self-report. Their 2019-mandate analysis covered 7.04 billion hours of playtime across more than 2.4 billion account profiles; a second dataset of 6.53 billion hours covered the 2021 mandate. Using an established definition of heavy play — more than four hours a day, six days a week — the 2021 dataset put heavy play at 0.44% of accounts before the restrictions and 0.59% after. Worth stating plainly, because it is easy to garble: the two hour totals belong to two separate analyses, and the 0.44/0.59 pair comes from the smaller one.

Different populations and definitions; not a like-for-like comparison. Sources: Zendle et al. 2023 (Nature Human Behaviour), 2021-mandate dataset; adolescent meta-analysis 2025; 16Best global play estimate of about 30.4 billion hours a week across 3.6 billion players, consistent with our gamer demographics statistics; National Centre for Gaming Disorders service evaluation, BJPsych Open 2024.

The Croatian data adds a nuance the headline rates erase, and it is the single cleanest demonstration on this page of why the instrument is the variable. Among those 1,239 players, endorsement of individual DSM-5 criteria ranged from 1.04% for damaged relationships up to 7.65% for playing to escape. Preoccupation, tolerance and escape sat at the top of the endorsement table; the negative-consequences items sat at the bottom. Yet when the authors modelled which criteria actually tracked poor mental health, the strongest links to depression, anxiety and low well-being came from damaged relationships, deception and escape, while tolerance, loss of control and withdrawal were the weakest.

Instruments that lean on engagement symptoms will catch enthusiasts. Instruments that lean on harm will not. Genre matters too, though less than the headlines imply: with role-playing games as the reference category, only sports and casual players scored significantly lower on IGD, while the remaining genres did not differ from each other. Broader context on how much people play and who they are sits in our gamer demographics statistics and mobile gaming statistics.

Myth 5: “Once you screen positive, you stay positive”

Reality: between a third and a half of people who meet the threshold still meet it a year or two later — the rest do not. A 2025 systematic review and meta-analysis of 50 longitudinal studies put categorical stability at 43–45% at one-year follow-up and 34–38% at two years. Symptom severity correlated moderately-to-strongly over time, so the underlying trait is not random; but the diagnostic label attaches and detaches far more than a cross-sectional survey implies.

None of this means the condition is fictional. In the Taiwanese clinical comparison by Yen and colleagues, among 60 patients diagnosed with gaming disorder, 96.7% reported health problems, 73.3% career impairment and 61.6% damaged social interaction, with 73.3% impaired across three or more domains at once. Where the disorder is real, it is severe and multi-domain. That study also found only 3.3% of patients in the action stage of change — the people who need help are largely not asking for it.

Who actually reaches treatment?

Very few people, and they look nothing like the average of the screening literature. The UK’s National Centre for Gaming Disorders, the country’s only specialist NHS service, opened in London in October 2019. Two published counts bracket its growth, and because both count the same thing — referrals of gamers plus family members — they can be compared directly.

Monthly rates are 16Best analysis, derived by differencing the two published cumulative totals. Sources: National Centre for Gaming Disorders service description, Journal of Behavioral Addictions 2022 (236 referrals to June 2021); NHS England, 28 March 2023 (745 referrals since opening). NHS England separately reported that the number of gamers treated rose by more than half from 2021 to 2022, and family members by 46%, but did not publish annual referral counts.

The patient profile is where the real information sits. A service evaluation published in BJPsych Open in 2024 profiled 380 of the gamers referred. Mean age 19, with 60% aged 13–18. 90% male, 9% female, 1% trans or other. Ten hours of gaming a day on average. One in ten with a formally diagnosed neurodevelopmental condition; one in eight with an existing mental health condition. Nearly half displayed aggression and 30% physical violence toward family when gaming was interrupted. Among the 17% making in-game purchases, average spending at referral was £4,500. The most common titles were Fortnite, Minecraft and Call of Duty — the most-played games generally, not a special category.

Did the national play-time laws work?

The two country-scale natural experiments produced weak or contested effects, and the country that pioneered the approach has since abandoned it. These are the only interventions at national scale with published outcomes, so they are worth dating precisely.

Sources: Choi et al. 2018 (Journal of Adolescent Health), Korea Youth Risk Behavior Web-based Survey 2011–2015; Zendle et al. 2023 (Nature Human Behaviour); Chinese adolescent survey cohort, N = 2,715, mean age 10.84, Journal of Behavioral Addictions 2025.

The Chinese row contains a genuine conflict, and it is instructive rather than embarrassing. Zendle’s telemetry sees accounts, not people: a minor playing on a parent’s or a rented account is invisible to it, and the study itself is careful to describe its finding as covering one segment of the industry. The 2025 survey cohort sees self-reported behaviour in 2,715 children with a mean age of 10.84 — young enough that circumvention is less likely, old enough to answer. Both can be right, and a third study shows the mechanism. A 2024 evaluation in the same journal followed 430 adolescent heavy gamers (mean age 12.75) and found 84.7% compliance. Look at how the remaining 15.3% broke the rules and the reconciliation appears: only 3% of the sample rented an account, but 14% played on a family member’s identity. That second route is exactly the behaviour telemetry cannot see. An adult account played by a child registers as an adult account playing more, not as a minor evading a curfew, which is one plausible route by which observed heavy play could rise while children’s own play falls. The same study found that 59% switched to short-form video and 51% to anime or TV — time displaced, not reclaimed.

Korea’s case is cleaner because the outcome is a repeal. Choi and colleagues tracked the policy across the Korea Youth Risk Behavior Web-based Survey from 2011 to 2015 and watched the effect die in slow motion: a 3.65-minute daily reduction in 2012, 3.20 in 2013, 1.14 in 2014, and by 2015 a 2.19-minute increase. Four years to go from a small effect to no effect to the wrong sign, with internet addiction scores and sleeping hours untouched throughout. The government announced abolition in August 2021, the National Assembly repealed the provisions that November, and a parental “choice permit” system replaced them. A decade-long clock-based intervention on the highest-profile gaming population on earth is the strongest available evidence that hours are the wrong lever, because hours were never the disorder.

Why do the sources disagree so much?

Because five separate methodological choices each move the answer, and almost no published article tells you which combination produced its headline. This is the section to keep.

• Instrument. The single biggest factor, and it is quantified: assessment-tool choice explained 77.97% of the variance across prevalence estimates in Stevens et al. (2021), against 20.4% for study region and 14.5% for year of publication. The Lemmens IGD-9, GAIT and PVP scales sit at the high end; the IGDT-10 sits low. Nobody has a gold standard — the 2025 BMC Psychiatry review states plainly that none exists. A detail that proves the point: Stevens et al. later issued a corrigendum removing two GAIT-based studies precisely because they had scored by percentile rather than the 5-of-9 cut-off. Even the meta-analysis of the instrument problem had an instrument problem.
• Criteria set. DSM-5 IGD (5 of 9, Section III, “further study”) versus ICD-11 GD (three behavioural criteria plus mandatory functional impairment). In one sample of 7,022, that choice alone produced 5.2% versus 2.7%.
• Symptoms versus impairment. Screening for symptoms catches people whose lives are intact. The criteria that predict depression are the harm-type ones; the criteria that inflate prevalence are the engagement-type ones. This is where over-counting is manufactured.
• Duration. ICD-11 asks for roughly 12 months of sustained pattern. A cross-sectional survey cannot observe duration, and the longitudinal evidence says only 34–45% of positives persist. Ignoring duration inflates the count by roughly 2.2–2.9× (16Best analysis).
• Sampling. Stratified random national samples yield 1.96%; the full literature yields 3.05%; self-selected gamer-only recruitment yields 8.1%. Where you find your respondents is a design decision with a 4× consequence.
• Denominator. “0.48% of the population” and “1.63% of players” are the same Croatian study. Quoting the second while implying the first — or the reverse — is the most common error in this field, and it is a 3.4× error.

Two further cautions. Measurement invariance is not established across cultures — the 2025 BMC Psychiatry review notes respondents from different origins rate items differently at identical underlying severity, which means cross-country comparisons of prevalence are on shakier ground than they look. And diagnostic instability is high: that review reports 60.4% of cases changed diagnostic status within a year, with “loss of interest” and “loss of control” the least stable items. When the country-level range inside a single meta-analysis runs from 1.2% (Norway, 2,055 participants) to 61.3% (Egypt, one study, 700 participants) — a 51× gap (16Best analysis), with pooled heterogeneity at I² = 100% — the first hypothesis should be instrument and sampling, not national character. For scale, China contributed 32 studies and 170,526 participants for a pooled 11.7%, and Spain five studies and 45,495 participants for 9.6%. The countries at the extremes are the ones with the thinnest evidence.

Key takeaways

• There is no single addiction rate. Published figures run from 0.48% to 8.6%, an 18× spread. Age explains a slice of that; method explains most of it.
• The instrument is the variable. 77.97% of between-study variance came from assessment-tool choice alone — against 20.4% for region and 14.5% for publication year. One sample of 7,022 gave 5.2% or 2.7% depending purely on which manual scored it.
• Engagement is not harm. In one Croatian sample, 7.65% endorsed playing to escape and 1.04% endorsed damaged relationships — a 7.4× gap between two items on the same questionnaire, and the harm items were the ones that tracked depression.
• Symptoms are not impairment. ICD-11 requires functional harm sustained about 12 months; most prevalence surveys test neither, and stability data says only a third to a half of positives persist.
• Hours are the wrong lever. Telemetry finds heavy play in under 1% of accounts; a decade of Korean curfew moved sleep and addiction scores not at all; China’s mandates show contested, partial effects.
• The clinical cases are real and severe. 96.7% of diagnosed patients report health problems and 73.3% are impaired in three or more life domains — while only 3.3% are actively seeking change.
• The treatment funnel is not the screening population. Clinic patients are 90% male and play about 8× the average weekly load; England and Wales’ only specialist service logged 745 referrals in three and a half years, against a lowest-case UK expectation above 300,000.
• So the spine, proven: when you read a video game addiction statistic, the first thing it tells you is which instrument was used. Ask for that before you accept the number — and if the article cannot tell you, it does not know what it measured.

Frequently asked questions

What percentage of gamers are addicted to video games?

There is no single figure. Published estimates range from 0.48% of a general population aged 15 to 64 (Croatia, IGDT-10, 2025) to 8.6% of adolescents (84-study pooled estimate, 2025), with the global pooled figure at 3.05% and 1.96% when only rigorously sampled studies are counted. Part of that range is genuine age difference, but the choice of assessment questionnaire alone explains about 77% of the variance between studies.

Is video game addiction a real medical diagnosis?

Yes, under ICD-11. The World Health Organization recognises gaming disorder as a condition requiring impaired control over gaming, escalating priority given to gaming, and continuation despite negative consequences, plus significant functional impairment normally evident for at least 12 months. The American Psychiatric Association lists Internet Gaming Disorder in DSM-5 Section III as a condition needing further study, not as a formal diagnosis.

How many people worldwide have gaming disorder?

Applying published rates to the global player base of about 3.6 billion gives a range of roughly 59 million to 310 million people, depending on which instrument and sample the rate came from. That 251-million-person gap is created by methodology, not by disagreement about the state of the world, which is why no responsible source should quote a single global figure.

Does playing a lot of video games mean you are addicted?

No. Play time and disorder are different variables. Telemetry covering 6.53 billion hours of play around China's 2021 mandate found heavy play, defined as over four hours a day for six days a week, in 0.44% of accounts before and 0.59% after, while symptom questionnaires report adolescent disorder rates about 15 times higher. Diagnosis requires functional impairment sustained over roughly 12 months, not a number of hours.

Who is most affected by gaming disorder?

Population screens find men about 2.5 times more likely to score positive than women, at 6.3% against 2.5%. Clinical populations skew far harder: the UK national gaming clinic reports 90% male and 9% female patients, mean age 19, with 60% aged 13 to 18 and average play of 10 hours a day. About one in ten had a diagnosed neurodevelopmental condition and one in eight an existing mental health condition.

Do gaming curfews and playtime limits reduce gaming addiction?

The evidence is weak and contested. South Korea's shutdown law, in force from November 2011, reduced adolescent internet use by about 3.65 minutes a day in 2012, an effect that had reversed to a 2.19-minute increase by 2015 with no change in addiction scores or sleeping hours; abolition was announced in August 2021 and the provisions were repealed that November. In China, telemetry found no aggregate reduction in heavy play after the 2019 and 2021 mandates, while a 2025 survey of 2,715 children with a mean age of 10.84 found 93.6% compliance and daily play falling from 60.1 to 43.5 minutes.

Do people recover from gaming disorder on their own?

Often, yes. A 2025 meta-analysis of 50 longitudinal studies found that only 43 to 45% of people meeting the threshold still met it one year later, and 34 to 38% at two years, with samples predominantly adolescent. Symptom severity remains moderately correlated over time, so the underlying tendency persists more than the diagnostic label does.

Sources

• World Health Organization — Addictive behaviours: gaming disorder (ICD-11 criteria and 12-month duration)
• Stevens, Dorstyn, Delfabbro & King (2021), Australian and New Zealand Journal of Psychiatry — Global prevalence of gaming disorder: a systematic review and meta-analysis
• Razum & Glavak-Tkalić (2025), Cyberpsychology — Prevalence and impact of Internet Gaming Disorder: a population-based study
• Borges et al. (2020), Canadian Journal of Psychiatry — (Internet) Gaming Disorder in DSM-5 and ICD-11
• Gisbert-Pérez et al. (2026), Addictive Behaviors — Prevalence of Internet gaming disorder in young adults: a systematic review and meta-analysis
• Burden of gaming disorder among adolescents (2025), systematic review and meta-analysis — Pooled prevalence 8.6% across 84 studies and 641,763 individuals
• Zendle, Flick, Gordon-Petrovskaya, Ballou, Xiao & Drachen (2023), Nature Human Behaviour — No evidence that Chinese playtime mandates reduced heavy gaming in one segment of the video games industry
• Choi et al. (2018), Journal of Adolescent Health — Effect of the online game shutdown policy on internet use, internet addiction and sleeping hours in Korean adolescents
• Temporal stability of problematic gaming and gaming disorder (2025), meta-analysis of 50 longitudinal studies — Categorical stability 43–45% at one year, 34–38% at two years
• Yen, Higuchi, Lin, Chou & Ko (2022), Journal of Behavioral Addictions — Functional impairment, insight and comparison between ICD-11 GD and DSM-5 IGD criteria
• National Centre for Gaming Disorders service evaluation (2024), BJPsych Open — Demographic profile and clinical characteristics of individuals accessing the service
• NHS England (28 March 2023) — NHS treats hundreds with gaming disorders
• Journal of Behavioral Addictions (2022), vol. 11 issue 2 — The National Centre for Gaming Disorders (UK): who is accessing this service? (236 referrals to June 2021)
• Compliance and alternative behaviours of adolescent heavy gamers under the Chinese online gaming restriction (2024), Journal of Behavioral Addictions — N = 430; 84.7% compliance, 3% rented accounts, 14% used a family member identity, 59% switched to short videos
• High compliance and reduced online gaming time among Chinese adolescents after the 2021 gaming policy (2025), Journal of Behavioral Addictions — N = 2,715, mean age 10.84; 93.6% compliance, 60.12 to 43.52 minutes a day
• Stevens et al. (2023), Australian and New Zealand Journal of Psychiatry — Corrigendum to the 2021 global prevalence meta-analysis
• Musetti, Floros, Chiappedi & Stavropoulos (2025), BMC Psychiatry — Gaming disorder in the ICD-11: the state of the game
• Przybylski, Weinstein & Murayama (2017), American Journal of Psychiatry — Internet Gaming Disorder: investigating the clinical relevance of a new phenomenon