TL;DR

About 2.8 million UK adults, or 5.1%, were paying for AI tools by June 2026, more than double the rate a year earlier, according to a working paper dated 5 November from Datta, Lambert and Schneider. Annual consumer spending has passed £1 billion. Because it rests on transaction records rather than surveys, this is a rare measure of what people pay for, and it shows a market still skewed towards the young, the well paid and London.

What the records show

The authors work from de-identified open banking and card data covering 14 million accounts, with support from the CAGE research centre, which is funded by the ESRC. That 5.1% compares with 2.3% twelve months before. Spending ran at £1.07 billion a year by July, on the project’s site.

The gaps are wide. Twenty-somethings are 5.9 times likelier to have paid than over-60s, a gap that narrows to 4.8 times once other factors are controlled for. In July the gender gap stood at 28% in men’s favour. Monthly earnings matter too: above £4,000, people are 88% likelier to have paid than those below £1,500. London leads the regions at 78.5 paying users in every 1,000 adults, with the South East next on 53.5.

The provider shift, with a caveat

The sharpest shift in the data is competition. Between December 2025 and July 2026, Claude went from 7% of direct subscription spending to 29%, while ChatGPT fell to 52.5%. The paper attributes part of that to technical and professional users on the most expensive plans.

That measure excludes purchases through app stores, so it is not a share of all users or even of all spending. Some press summaries date the 5.1% to July; the authors’ site and paper give June. Readers should also note that this is a working paper, and that it measures consumer payments, so it does not settle how much AI use reaches people through work.

Looking Forward

The government’s own adoption survey covers businesses, so this offers the household side of the same picture. The authors publish a new release each month, which gives UK analysts a regular series to follow. In our view, the concentration matters more than the total: if paid tools do raise productivity, a market led by younger and better-paid Londoners could widen existing divides, and that would be a question for skills policy, not just for the AI companies competing over the spending.