%201.webp)
A study by economists at the University of Warwick and Imperial College London found that 5.1% of UK adults paid for AI assistants in June 2026, equivalent to about 2.8 million people.
Based on anonymised bank and card transactions, the analysis found UK consumer spending on AI tools has grown from almost nothing in early 2023 to an annual market worth GBP £1 billion. It tracks personal payments for standalone AI products, including subscriptions bought through mobile app stores, but excludes free use and licences paid directly by employers.
The figures point to rapid growth in a market led by consumer subscriptions to services such as ChatGPT and Claude. Subscriber numbers have roughly doubled each year, suggesting paid AI use is moving beyond a niche group of early adopters.
Work patterns
Occupation was one of the clearest indicators of who pays for AI out of pocket. Subscription rates were estimated at about 20% among IT professionals, 13% to 16% among university lecturers and finance professionals, and 7% to 9% among school teachers, writers, artists and doctors.
At the other end of the scale, fewer than 3% of electricians, nurses and social care workers were estimated to pay for AI tools themselves. This suggests paid use is concentrated in jobs where workers see a direct benefit in day-to-day tasks.
"Millions of people are now putting their hands into their own pockets to pay for AI every month. In little more than three years, these tools have become a billion-pound consumer market in the UK. This surge has been phenomenal and with only one in twenty adults currently subscribing, there's still room for further growth. The big question is how many people currently using free AI tools will eventually decide they are worth paying for?" said Dr Peter Lambert, Department of Economics, University of Warwick.
Another researcher said the occupational split may reflect how people use these products. "People whose jobs involve tasks that AI can help with are more likely to pay for it themselves, so that suggests people are buying these tools partly for what they can do for them at work. The first paying customers were disproportionately men and high earners, but that picture is changing rapidly, with women now making up nearly half of new subscribers," said Dr Nikhil Datta, Department of Economics, University of Warwick.
Regional split
London recorded the highest share of paid AI users, with 7.9% of adults estimated to have subscribed in June 2026, compared with the national figure of 5.1%.
The North East ranked lowest among UK regions at 3.8%. The gap suggests paid AI adoption is stronger in the capital, in line with broader patterns of digital spending and the concentration of professional and technical work.
Market shares
ChatGPT remained the largest brand by subscription spending during the period covered by the study, but Claude gained ground quickly. Claude's share of subscription spending rose from 7% to 29% between December 2025 and July 2026, while ChatGPT's share fell from 74% to 53%.
Claude's growth reflects greater uptake of more expensive plans, especially among people in technical and professional roles. It had stronger support among software and IT workers, engineers, finance professionals and academics, while teachers, healthcare workers and human resources staff tended to favour ChatGPT.
"Like many early-stage markets, the picture can change very quickly. ChatGPT still leads at present, but its early dominance is shrinking. Claude captured almost a third of consumer subscription spending by July, up from just 7% seven months earlier," said Dr Patrick Schneider, Imperial Business School.
Demographic trends
The study also found a sharp age divide. Adults under 30 were five times more likely to pay for AI than those over 60.
Men were 27% more likely to subscribe than women, while people in the top fifth of earners were almost twice as likely to pay as those in the bottom fifth. Even so, the customer base is broadening as adoption spreads beyond the higher-income male users who dominated the market in its earlier stages.
The study was written by Nikhil Datta and Peter John Lambert, both affiliated with the University of Warwick and the Centre for Economic Performance at the London School of Economics, and Patrick Schneider of Imperial Business School.