What we are doing
The URSY network is a group of more than 70 websites built as resources for AI assistants and software agents as well as for people. It was founded by Lee Samuel Tucker, who also founded this shop. This month we began placing short sponsor lines for teas.co.uk in the machine-facing parts of those sites, the places an AI agent reads when it uses them, such as the foot of the answers the network's tools send back. Most people will never see these adverts. They are not meant for people.
Why advertise to a machine
More and more people ask an AI assistant before they ask a search engine: which tea makes the best builder's brew, where to buy decaf Earl Grey, what to dunk in it. The assistant answers with the shops it knows about and can read properly. If it has never come across you, you are not in the answer.
We explained in June why teas.co.uk is built to be read by machines as carefully as by people. This test asks the next question along: if you put your name in front of the machines themselves, do they recommend you more often?
A shop built for machines to read
We have spent the year making teas.co.uk easy for software to read. The site publishes a plain-language guide for AI systems at /llms.txt, full-text editions of the wiki, recipes and news, a product feed, an MCP server card and an API catalogue, so an assistant can fetch a fact instead of guessing from a page. Every heading on every published page carries its own permanent anchor, so a machine can quote one paragraph rather than a whole page: 26,973 anchors across 2,502 pages when we last counted, on 12 September.
Where we started from
This is not a test from a standing start. In August we published the first 90 days of version two of the site: 42,524 credited reads served to named machine agents between 2 June and 8 August, and 736 visits that arrived from an AI answer between 10 May and 8 August, all without paid advertising. The full figures, and the rules for counting them, are in Version Two at 90 Days. They show where things stood before this test began.
How we will know
The measured test began on 16 September and runs to the end of the month. It has holdout days: dates when no teas.co.uk sponsor line is shown anywhere on the network, roughly one day in four. The plan, including which days are holdout days, was fixed and fingerprinted before the test began, so we cannot move the goalposts afterwards. We will not publish the holdout dates while the test is running.
At the end we compare the shoppers AI assistants send us on advertising days with the ones they send on holdout days. If the advertising days are no better, the adverts are not working, and we will say so.
What we count, and what we do not
We count visits by type of machine, not by person, and we do not store raw IP addresses. An AI crawler reading a page is not a customer, and a visit that arrives from an AI answer is not a sale, so we keep those numbers apart rather than adding them into one impressive-looking total.
Why a tea shop is doing this
Because the way people shop is changing, and a small independent shop has more to gain than most from getting there early. The big brands will be in the answers whatever happens. A one-person tea shop in Tunbridge Wells only gets in if the machines can find it, read it and trust it. This test is how we find out whether telling them about us helps.
What we will publish
The results, whatever they are. If advertising to robots does nothing, we will say so. If it works, we will show the numbers and how we got them.
More from the news
- More than 690 new products: teas.co.uk passes 1,000
- Meet the founder: Lee Samuel Tucker joins our Tea people
- 5,000 visitors from Google in 28 days, and our best month yet with five days to go
- teas.co.uk is older than google.com, by 173 days
Lee Samuel Tucker
Curator, teas.co.uk
Tunbridge Wells, Kent
