Claude’s UK AI Subscription Spending Share Quadruples as Market Grows

The Who Pays for AI? study, published November 5, 2026, found that Claude’s share of directly billed UK consumer AI subscription spending quadrupled from 7.4% in December 2025 to 29.4% in July 2026.

The paper by Nikhil Datta, Peter John Lambert and Patrick Schneider estimated that 5.1% of UK adults, about 2.8 million people, held an active paid AI subscription in June 2026, up from 2.3% in June 2025. The July 2026 dashboard recorded 52.0 paid users per 1,000 adults and monthly spending of £1,600 per 1,000 adults, equivalent to a £1.07 billion annual rate across UK adults. That figure annualizes July’s rate rather than totaling spending for a completed year.

The authors attribute 88% of spending growth between January 2024 and July 2026 to more people paying and 12% to higher average spending per subscriber. Average direct monthly spending per active subscriber rose from £22 to £28 over the same period. The market’s expansion therefore comes mainly from a widening payer base, with upgrades and multiple subscriptions contributing a smaller share.

Claude’s Revenue Share Outpaces Its Subscriber Base

The study’s provider comparison shows ChatGPT’s directly billed spending share falling from 73.5% in December 2025 to 52.5% in July 2026, while Gemini accounted for 5.1% and other products for 13.0% in July. ChatGPT retained 71.2% of active general-purpose assistant subscribers, compared with Claude’s 27.1%. Because the total market was growing, both companies’ revenues grew despite their changing shares.

The authors’ pricing analysis explains why Claude’s spending share exceeds its subscriber share. By mid-2026, 6.2% of active Claude subscribers paid at least £80 per month, compared with 0.3% of ChatGPT subscribers. Standard-tier prices remained around £19 per month on average, while the providers added plans above and below that tier. About a quarter of Claude’s direct spending-share gain came from power-user upgrades, with the remainder from new users.

The authors’ switching and release analyses show how the customer base changed. Among ChatGPT subscribers who cancelled, 1.4% started paid Claude within 60 days throughout 2025; between March and May 2026, the rate reached 11.7%. Only 4.2% of assistant subscribers paid for two or more assistants in the same month. Major Anthropic releases were associated with a 39.5% increase in first-time paid sign-ups over 14 days, against 10.0% on random comparison dates; the analysis does not establish that releases alone caused the increase.

Adoption Remains Concentrated by Job and Age

The researchers’ demographic estimates show paid adoption remains uneven. Cumulative paid adoption through July 2026 was 17.9% among adults aged 20–29 and 3.0% among those 60 and over; after adjustment for income, sex and region, the corresponding figures were 18.3% and 3.8%. July also recorded 62.2 active paid AI users per 1,000 men and 48.6 per 1,000 women, a 27% adoption gap, while average monthly spending was £26 for men and £21 for women.

Income alone does not explain the divide. Cumulative adoption was 10.0% among workers taking home under £1,500 per month and 18.8% among those taking home over £4,000, but the authors estimate that occupation, industry, location and demographics explain 64% of the raw income relationship. Direct paid adoption in June 2026 was highest among IT professionals at 20.3%, university and college lecturers at 16.0%, and finance and insurance professionals at 15.8%, compared with 2.5% for nurses and midwives and 1.7% for social care workers.

The paper maps a related geographic divide: active paid adoption in London was 6.8% in June 2026, nearly twice the 3.6% rate outside London, with clusters in Cambridge, Oxford and Edinburgh.

Bank Records Reveal Payments and Measurement Limits

The project’s methodology uses de-identified records from 14.2 million UK bank and card accounts and about 8 million transactions a week between January 2022 and August 2026. The headline national analysis used 457,072 active main accounts in June 2026 after activity and spending screens. Direct payments are matched to a registry of more than 200 standalone AI products using statement text, exact plan prices and launch dates. Pre-existing software suites such as Microsoft 365, Canva and Adobe are excluded.

Apple and Google app-store charges do not identify the AI product, so subscriptions are inferred from recurring charges at AI price points and account characteristics. Approximately two-fifths of estimated active subscriptions come through these concealed channels. Provider spending shares cover direct payments only; they also exclude accounts where privacy filters obscure Claude’s name and re-weight the remaining accounts. Monthly weights match official age, sex, region, income and neighbourhood distributions, scaling the estimates to 55.5 million UK adults aged 18 and over.

The study covers subscriptions people buy themselves and excludes employer-purchased licences. It counts accounts rather than directly observing unique people. The authors’ sensitivity analysis places individual adult payers between 4.4% and 6.0% when the account-to-person conversion assumption varies by 15% in either direction.

The records separate subscriber leadership from spending leadership: ChatGPT retains the larger assistant subscriber base, while Claude’s gains are tied to higher-priced plans, new users and switching. With most market growth coming from new payers, the central open question is whether paid adoption broadens beyond its current occupational and demographic concentrations or continues to depend on the groups already driving subscription spending.