There is no single number, and anyone who gives you one without naming a survey is guessing. The World Health Organization estimates that 1.2% of the world’s adults have a gambling disorder. National surveys put the share of adults scoring in the top problem-gambling band anywhere from about 0.4% to 2.6%.
Most of that spread is not a real difference in gambling. It is a difference in how the surveys were run. Two official surveys of the same country, published months apart, can differ by a factor of six — and both can be correct about what they measured.
This page sets out the main figures, states the survey, year, instrument and mode behind each one, and explains why the numbers disagree. If you want to know how many people have a gambling problem, that context is the answer, not a footnote to it.
The short answer
- Globally: WHO estimates 1.2% of adults have a gambling disorder.
- Great Britain: 2.4% of adults scored 8 or more on the PGSI in the Gambling Survey for Great Britain 2025 — but 0.4% of adults in England did so in the Adult Psychiatric Morbidity Survey 2023/24.
- Australia: 2.6% of adults were in the high-risk PGSI category in the 2024 national pilot study.
- United States: no comparable national PGSI prevalence figure exists. The best-known national survey measures something different.
None of these is a count of people with a diagnosis. That distinction does a lot of work, so it is worth understanding first.
What these surveys actually measure
Nearly every national figure you will see comes from the Problem Gambling Severity Index, or PGSI. It is nine questions about the past twelve months, scored from 0 to 27 and reported in bands:
| PGSI score | Usual label |
|---|---|
| 0 | No problems reported |
| 1 to 2 | Low risk |
| 3 to 7 | Moderate risk |
| 8 or more | High risk, often labelled “problem gambling” |
The PGSI is a screening tool. It is not a diagnosis, and it is not administered by a clinician. The Gambling Commission’s own guidance on using its statistics lists “as a measure of addiction to gambling” among the things the figures should not be used for, and says the PGSI “should not be confused with a measure of gambling addiction”.
A clinical diagnosis is a different thing entirely. Gambling disorder is defined in the DSM-5-TR and in the WHO’s ICD-11 as code 6C50, and it is made by a clinician assessing one person, not by a questionnaire scored in a spreadsheet. Our explainer on what problem gambling means sets out how the screening tools and the diagnostic manuals fit together.
So when a headline converts “2.4% scored 8+ on a screening questionnaire” into “1.3 million gambling addicts”, it has invented a claim the data cannot support.
The global estimate
The WHO gambling fact sheet, published on 2 December 2024, says that “estimates suggest that 1.2% of the world’s adult population has a gambling disorder”.
The same fact sheet gives a wider figure for harm: “Around 5.5% of women and 11.9% of men globally experience some level of harm from gambling.” WHO attributes that to the Lancet Public Health Commission on gambling. It also reports that “for every person who gambles at high-risk levels, an average of six others (usually non-gamblers) are affected”, and that people gambling at harmful levels generate around 60% of gambling losses.
Those last two points matter for how you read every number below. The people who score 8 or more are a minority of those affected, and they are not a minority of the money. Our page on global gambling harm in numbers goes through the international picture in more detail.
The national figures, with their small print
Read this table across, not down. The instrument and mode columns are the reason the percentages differ.
| Survey | Population and period | Instrument | Mode | Headline figure |
|---|---|---|---|---|
| Gambling Survey for Great Britain, Annual Report 2025 | Adults in Great Britain | PGSI, past 12 months | Push-to-web self-completion, with a postal option; around 20,000 respondents a year | 2.4% scored 8+ (95% CI 1.9–2.8); 3.5% scored 3–7; 7.8% scored 1–2 |
| Adult Psychiatric Morbidity Survey 2023/24, published 27 November 2025 | Adults in England, fieldwork March 2023 to July 2024 | PGSI, past 12 months | Self-completion section inside an interviewer-recruited in-home interview | 0.4% scored 8+; 1.6% scored 3+ |
| National Gambling Prevalence Study Pilot 2024 (AGRC) | Australian adults, August–October 2024, n=3,881 | PGSI, past 12 months | Probability-based mixed mode: random digit dialling, address-based sampling and a probability online panel; completed online or by telephone interview | 2.6% high risk; 4.8% moderate risk; 7.6% low risk; 65.1% gambled |
| NGAGE 3.0, fieldwork January–March 2024, n=3,013 | US adults | Four self-reported problematic gambling behaviours — not the PGSI | Online survey | 8% reported at least one indicator “many times” in the past year, down from 11% in 2021 |
Sources: the Gambling Commission’s GSGB Annual Report 2025 and its survey methodology page; NHS England’s APMS 2023/24 chapter on gambling behaviour; the Australian Institute of Family Studies’ National Gambling Prevalence Study Pilot 2024; and the National Council on Problem Gambling’s NGAGE 3.0 key findings.
Two of these figures deserve extra care. The GSGB also reports the same measure with a different base: among adults who had gambled in the past 12 months, 4.0% scored 8 or more, against 2.4% of all adults. Same survey, same question, different denominator. And the American 8% is not a problem-gambling prevalence rate at all — the NCPG’s own summary describes people reporting a problematic behaviour “many times”, which is a lower bar than a PGSI score of 8.
Why Britain has two official numbers six times apart
This is the single most misreported thing in gambling statistics, so it is worth going slowly.
The GSGB says 2.4% of adults score 8 or more. The APMS says 0.4%. Both are official government statistics. Both used the PGSI. Neither is lying.
It is mostly about who takes part
The Gambling Commission commissioned Professor Patrick Sturgis to run experiments on exactly this question. Its summary of that research, published on 14 August 2025, found two effects:
- Who agrees to take part. When the survey invitation mentioned gambling, reported gambling participation rose by about 4 percentage points. People who gamble are more likely to open an envelope about gambling.
- Whether someone is listening. The share scoring 1 or more on the PGSI was 4.4 percentage points higher in online self-completion than in telephone interviews, which points to people under-reporting to a human being.
A common myth about the APMS
You will often read that the APMS number is lower because an interviewer asked the gambling questions out loud. That is wrong. The APMS gambling questions sat in the self-completion section of the phase-one interview — the respondent answered them privately.
What differs is everything around that moment: the APMS recruits people at home, in person, for a long survey about mental health in general, while the GSGB recruits by post for a survey about gambling, completed online. Different people agree to take part, in a different frame of mind, having been told a different thing about the subject. That is the dominant mechanism, not the presence of an interviewer at the moment the question is asked.
What the statistics regulator said
The Office for Statistics Regulation published its review of the GSGB on 22 May 2025. It noted Professor Sturgis’s conclusion that there is “a non-negligible risk that GSGB statistics over-state the true level of gambling and gambling harm in the population”, and criticised the Gambling Commission for not putting that caveat in the statistical releases themselves. The GSGB figures are official statistics, not accredited official statistics.
The Commission’s guidance is equally blunt in the other direction: GSGB results should not be used “to provide direct comparisons with results from other gambling or health surveys”, and “only comparisons with previous waves of the GSGB are appropriate when examining changes over time”.
Put those together and you get the honest position. The two British numbers bracket a range. They do not average out to a true value, and neither one of them can be quoted as the answer.
Is it getting worse?
Within a single survey, where comparison is fair, the answer for Great Britain is mostly no. The Gambling Commission’s GSGB trends report covering 2023 to 2025, published on 16 July 2026, concludes that “overall, there were few differences in specific gambling behaviours examined between 2023 and 2025”.
The share of respondents scoring 0 on the PGSI moved from 76.1% in 2023 to 76.9% in 2025. Past-four-week gambling participation went from 48% to 47%. The report does flag a rise in men gambling on activities other than lotteries, and a rise in 35 to 54 year olds seeking help for their own gambling.
That is the value of a consistent survey run the same way every year. It cannot tell you the true level, but it can tell you the direction — which is exactly what a set of incompatible one-off surveys cannot do.
How precise are these percentages?
Less precise than they look. The GSGB 2025 figure of 2.4% carries a 95% confidence interval of 1.9% to 2.8%, meaning the survey’s own best estimate spans nearly a full percentage point. On a population the size of Great Britain, that range covers several hundred thousand people.
Confidence intervals matter most when someone claims a small year-on-year movement is a real change. If two intervals overlap heavily, the sensible reading is that nothing measurable happened.
The instruments themselves add another layer. The Gambling Commission’s page on problem gambling screens explains that the PGSI runs to nine items scored 0 to 27, while the older DSM-IV screen uses ten criteria with a threshold of three, and warns that estimates from the two screens “should not be combined”. Some national surveys still report against DSM-based criteria rather than the PGSI, so their headline rates are not interchangeable with a PGSI 8 or more figure from another country, even when both are described as "problem gambling".
Four mistakes to avoid when you read a gambling statistic
- Turning screening scores into addicts. A PGSI score is a risk signal, not a diagnosis, and the regulator says so in writing.
- Comparing across surveys. Two figures collected by different methods are not a trend. A rise from 0.4% to 2.4% between two different surveys is an artefact of method, not an epidemic.
- Ignoring the base. “Of all adults” and “of people who gambled” are different populations. In the GSGB 2025 the same measure reads 2.4% or 4.0% depending on which you pick.
- Mixing instruments. A PGSI 8+ rate, a DSM-5-based rate and a count of self-reported behaviours are three different measurements. Lining them up in one chart produces nonsense.
The number you want is probably not the top band anyway
If you are asking how many people have a gambling problem because you want to know how big the problem is, the highest PGSI band is the wrong measure to fix on.
In the GSGB 2025, 2.4% of adults scored 8 or more — but another 3.5% scored 3 to 7 and 7.8% scored 1 to 2. In the Australian pilot, 2.6% were high risk while a further 4.8% were moderate risk and 7.6% low risk. Because the lower bands contain far more people, a large share of all gambling harm is experienced by people who would never be counted in the headline figure.
Add the WHO estimate that six other people are affected for every high-risk gambler, and the population touched by gambling harm is an order of magnitude larger than any prevalence percentage suggests. That is the case for treating gambling as a public health issue rather than a small clinical one.
Why the exact number matters so much
Prevalence figures are not academic trivia. They decide how much money goes into treatment, how urgent a regulator looks, and how a gambling industry argues its case in public. A 0.4% figure describes a niche clinical problem. A 2.4% figure describes a population-level one. The same country, in the same period, produced both.
That is why the methodology is not a technicality to skip past. Anyone quoting one of these numbers without saying which survey it came from, and how, is either careless or selling something.
If you are asking because of your own gambling
Prevalence rates say nothing useful about one person. Whether your gambling is a problem is a question about your money, your time and your relationships, not about where you sit in a national percentage.
If the question is on your mind at all, our guides to how gambling addiction develops and how to get help are more use than any of the figures on this page. Support is free and confidential in most countries, and you do not need to score 8 on anything to deserve it.
In short
How many people have a gambling problem? Worldwide, WHO estimates 1.2% of adults have a gambling disorder. In wealthy countries with good surveys, between roughly 0.4% and 2.6% of adults score in the highest problem-gambling band, and the method used explains most of that gap. Several times that number experience some harm, and around six people are affected for every one person gambling at high risk.
Any figure quoted without its survey, year, instrument and mode should be treated as an opinion. How we check and correct what appears here is set out in our editorial policy.
Last reviewed: 30 September 2026. Sources checked on this date.