AI visibility for UK ecommerce: the VAT inclusive price is the number an assistant repeats
Measured on 15 October 2026 across 204 UK readings of ChatGPT and Gemini: why a British product page carries two prices, and the delivery, returns and specification detail that decides the answer.
Published by AI Knows Us · Updated 16 October 2026
A British online shop wins an assistant's answer on one fact above all others: the price a UK shopper actually pays, in pounds, inclusive of VAT, written in text on the page and matching the markup underneath it. A British product page usually carries two prices, one with VAT and one without, and an assistant repeats whichever one it finds first. Everything else a retailer controls sits behind that: delivery written as a UK shopper would ask about it, the returns position in words rather than in a PDF, and a specification in text rather than in an image. The UK half of our 15 October 2026 benchmark was 204 readings, 102 questions put to ChatGPT and the same 102 put to Gemini, and our own domain came back once, from ChatGPT, and never from Gemini.
What the measurement covers, stated plainly
That run asked buying questions in one category, AI visibility, in three markets. It did not run a separate measurement for retail, so there is no shopping citation rate in it and nobody should quote one. What it does carry is the source mix, which is the part a retailer can act on, because it says where citations go rather than which shop won.
| Figure | Value |
|---|---|
| Date run | 15 October 2026 |
| Engines | ChatGPT and Gemini |
| Usable readings | 597 of 612 attempted |
| UK readings | 204 |
| UK readings citing our own domain | 1 |
| Citations recorded | 10,652 across 2,708 distinct domains |
| Vendor's or agency's own site | 85.6% |
| News or a large publisher | 2.5% |
| Review directories | 0.8% |
Our product reads five surfaces, ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. This dataset covers two of them on one date, and the method, the market comparison and the full source mix are set out in what AI assistants actually cite across 597 readings of ChatGPT and Gemini in three markets.
The two prices problem, which is a British problem
A UK retailer selling to consumers quotes a total price including VAT. A UK retailer selling to trade quotes the price excluding it, and often shows both. On the page that is clear enough to a human reading the small print. To a fetcher it is two numbers near each other with no instruction about which one is the answer, and the mismatch gets worse when the structured data carries one and the visible page shows the other.
- Put the price a shopper pays in the HTML, in pounds. If the figure is injected by a script after render, treat it as a figure that may never be read.
- Make the markup agree with the visible price. Google asks that structured data match what the user sees, so a price in markup that the page does not show is a defect rather than an advantage. The property level detail is in Google product schema and the mismatch that breaks it.
- Label the trade price as the trade price. Where both are shown, say in words which one excludes VAT, next to the figure rather than in a footer.
- Set the currency explicitly. A bare number reads as the shop's home currency, and for a retailer selling into several markets that is how a British shopper gets quoted a figure they cannot buy at.
Stale prices are the other half of this. When the feed, the marketplace listing and the page disagree, the oldest one is often the one repeated back at you, and the diagnosis is in why AI quotes an old price.
Delivery and returns are the questions, so answer them in text
British shoppers ask about delivery and returns more precisely than most product pages answer. Free UK delivery is a claim with exceptions behind it, and the exceptions are the part a buyer in Inverness, Stornoway or Belfast is asking about. Write them out.
| What a UK shopper asks | What belongs on the page in text |
|---|---|
| Will you deliver to me | Mainland, Northern Ireland, the Scottish Highlands and the islands, each named, with the charge for each |
| When will it arrive | The cut off time, the working days, and what a bank holiday does to it |
| What if I change my mind | The cancellation right for online orders under the Consumer Contracts Regulations, and who pays return postage |
| What if it arrives faulty | The right to reject faulty goods under the Consumer Rights Act 2015, and the route to use it |
| Is the saving real | A was and now claim needs the higher price to have genuinely been charged, which is a CAP Code matter the ASA can be asked about |
None of that is new law to a British retailer. The point is where it lives: a policy page nobody links to, or an accordion that loads on click, while an assistant reaches for whatever is in text on the page in front of it.
Specification in text, which is also an accessibility job
Dimensions, materials, compatibility, care instructions and sizing belong in a list on the page. A specification inside a JPEG or a PDF cannot be quoted, and it also cannot be read aloud, which is the same defect viewed from the Equality Act 2010 duty to make reasonable adjustments for disabled customers. Treating the two as one job is the cheapest way to get the work funded, because the accessibility case and the visibility case point at the same change.
Sizing deserves its own note. A size guide giving UK sizes alongside the measurement, in text, answers a question that currently sends British shoppers to a marketplace listing instead.
Where the budget should not go, on the evidence
The instinctive answer to being absent from an AI shopping answer is to buy presence on review platforms and aggregator pages. Review directories took 0.8% of the 10,652 citations recorded on 15 October 2026, and news or large publisher pages took 2.5%, while 85.6% went to a page on the vendor's or the agency's own site. That is a strong argument for spending the quarter on your own category and product pages, and for treating the editorial roundups as a slower, separate piece of work rather than the first purchase.
Category pages carry the most of this weight, because shoppers ask for a category with a constraint far more often than they ask for a product by name, and most British category pages are a grid of tiles with a sentence of text. If you are deciding whether to do this in house or buy it, the tiers and what each one charges are set out in what each tier of SEO company charges. The platform level version of the product data work, written without the British grain, is in AI visibility for ecommerce.
One more British field: the email capture
The discount box on a British shop runs on UK GDPR and PECR, not on American marketing rules, and the practical difference is the soft opt in for people who have already bought from you against consent for everybody else. It matters here for one reason: a shop that states its consent position in plain words on the page is publishing another verifiable fact about itself, and verifiable facts are what an assistant can commit to in an answer.
Questions about AI visibility for a UK online shop
Which price should the structured data carry?
The one on the page. If the page shows the consumer price including VAT, the markup carries that figure. Matching the two matters more than which of the two you chose.
Do I need to be on marketplaces to be recommended?
A marketplace listing is worth keeping accurate because it is another place your current price and stock appear. The measured source mix says the bigger lever is your own pages, with 85.6% of 10,652 citations going to the brand's or vendor's own site on 15 October 2026.
Does a .co.uk domain help a British retailer get cited?
What showed up in the UK readings was British specificity rather than a particular domain ending: pages written about the British version of the question, by companies nobody would call large. A .com with pound pricing, UK delivery detail and British wording carries the same signal.
How often should a product page be re-checked?
Monthly for price and stock, and after every feed change, so the figure in the answer is the figure you are charging. Ask the shopper questions again on the same schedule and record who was named.
What about Northern Ireland specifically?
Name it. A delivery page that says UK and means Great Britain produces an answer that is wrong for a shopper in Belfast, and that is a correction you will spend far longer undoing than writing the line in the first place.
The shop that answers its own questions
Most British product pages are already carrying the facts an assistant needs. They are carrying them in an image, in an accordion, in a feed that disagrees with the page, or in two prices with nothing to say which is which. Fixing that is a week of unglamorous work on pages that already exist, and it is measurable: ask the shopper questions, record what gets named, change the pages, ask again. We sell that measurement and the pages that follow it, so the honest version of our own position belongs here too, which is one citation of our domain in 204 UK readings on 15 October 2026. AI Knows Us was built to close that gap for other people's catalogues as well as our own.