Mastercard forecasts shopping agents for one in ten customers — the homework for merchants is called product data
On 8 September 2026 Mastercard published the report A Short History of the Future of Shopping and Payments. The forecast: by 2030 more than one in ten online shoppers will regularly have an AI agent do the buying, over 300 million people worldwide. The concrete demand on merchants is unspectacular and still the real news: product information, advertising claims and policies have to become machine-readable and verifiable.
What happened
On 8 September 2026 Mastercard presented the report “A Short History of the Future of Shopping and Payments” in London. The core forecast: by 2030 more than one in ten online shoppers will regularly have an AI agent buy on their behalf, over 300 million people worldwide; Mastercard puts the associated demand volume at £370 billion a year. It is based on a survey of 13,000 parent-teenager pairs in 13 markets — including Germany, France, Italy, Spain, the Netherlands and the United Kingdom — plus contributions from four futurists. The individual usage shares for parents and teenagers are reported differently by the trade publications, and the original release was not retrievable for this assessment; they are therefore not given here. The report becomes concrete on what it requires of merchants: product information, advertising claims and policies would have to become machine-readable — a sustainability claim, for instance, backed by a certificate that an agent can find, read and verify. The commissioning party and sender is a card scheme that earns money from agentic payments; that belongs in any evaluation of the forecast.
Who it affects
Online merchants of every size, marketplace sellers, manufacturers with direct sales, plus everyone who maintains product data — PIM owners, shop managers, agencies. For bricks-and-mortar retailers without an online business it is a footnote for now.
Assessment
Forecasts out to 2030 are marketing, no matter who they come from, and a figure in pounds for a global market does not make them any more solid. The report is still interesting, because the homework in it has nothing to do with the future: clean, structured, verifiable product data. Anyone who today delivers prices, availability, shipping costs, return periods and product attributes only as a pretty page and not as readable data will not be rated wrongly by an agent — they will simply not be found. It is the same mechanism merchants know from fifteen years of price comparison engines, only stricter, because an agent also reads conditions and not just prices.
The second half of the task is regularly missing from such reports, and it is the more expensive one: liability. An agent that selects and pays generates chargebacks, complaints and disputes for which there are still no solid rules. Who carries the chargeback when an agent buys the wrong thing on the customer’s behalf is written in no rulebook a German merchant is handed today. My order of priorities for merchants is therefore clear: first get master data and return conditions in order — that pays off in conversion and fewer returns regardless of any agent forecast — and only then talk about agent access in the checkout, and then with questions about liability rather than with enthusiasm.
What to do now
- Check your own shop: are price, availability, shipping costs, delivery time and return period retrievable as structured data, or do they only appear in body text and images?
- Make advertising claims provable. Every assertion that sells — origin, certification, warranty — needs a linked, verifiable proof point. That helps today under competition law and later with the agent.
- Ask your payment service provider how it flags agent-initiated payments and who is liable in a dispute. Without that answer, no agent belongs in your checkout.