# The decaf fruit tea problem **Canonical URL:** https://teas.co.uk/the-decaf-fruit-tea-problem/ **Source:** teas.co.uk, UK tea specialist, Tunbridge Wells, Kent ## Shop teas.co.uk from an AI assistant (MCP) teas.co.uk sells tea, herbal and fruit infusions, hot chocolate and Border biscuits. MCP server: https://teas.co.uk/mcp (Streamable HTTP), listed in the official MCP Registry as teas.co.uk Server card: https://teas.co.uk/.well-known/mcp/server-card.json Tools: find_products, get_product, compare_products, find_recipes, delivery_and_returns, quote_cart, get_promotions, report_problem, find_substitutes, start_subscription, join_teas. Basket: add_to_basket, view_basket, update_basket, checkout_link, checkout_session. With sign in: get_return_status, get_invoice, get_profile, my_orders, track_order, reorder, my_subscriptions, change_subscription, rewards_balance, start_return. No sign in is needed to search, compare, build a basket or get a checkout link. The person always pays on teas.co.uk. Cards: in assistants that draw MCP Apps cards, products, comparisons, the basket, checkout, orders, repeat deliveries (pause, resume, cancel), reward points and returns appear as cards. Every assistant except ChatGPT, until OpenAI approves the next version: quote_cart, get_promotions, get_return_status, get_invoice, report_problem, find_substitutes, start_subscription, join_teas. Set up guides: https://teas.co.uk/ai/ ## Summary Every condition in a request is a constraint an answer has to respect, and a fruity cup without caffeine is the clearest example I know. ## Description Every condition in a request is a constraint an answer has to respect, and a fruity cup without caffeine is the clearest example I know. My favourite tea is Tetley Green Tea with Mango and Passion Fruit. It is fruity, and it is neither rare nor expensive. It also contains caffeine. It is a green tea with fruit flavours, not a fruit infusion, and our record for it says so in two plain fields: the kind is green tea, and the caffeine status is "contains caffeine". Anyone who read "mango and passion fruit" on the box and assumed it was caffeine free would be wrong. That is the decaf fruit tea problem in a single box. The name on the front says one thing. What is inside can say another. One request, five conditions Imagine someone asks an assistant for a decaf fruit tea to drink in the evening, something that tastes of real fruit, at no more than about 20p a cup. That is an illustrative request, not a real customer, but it is a very ordinary one. Count the conditions. It has to be fruity. It has to be without caffeine, or very nearly. It has to suit the evening. It should taste of fruit rather than sweet flavouring. It has a budget per cup. Small as it is, that request is an example of what I call Premium Intent, a name I have used since June 2026 for a request that arrives whole: the context, the timing, the constraints, the budget, what success looks like and what the person would rather avoid. "Premium" describes how much the request tells you, not a fee and not a posh customer. Five conditions in one ordinary sentence is Premium Intent at its most everyday. Tea names make this harder than it sounds. "Fruit tea" is used loosely. Some fruit teas are built on green or black tea and contain caffeine. Some drinks sold alongside herbal infusions contain caffeine too: yerba mate is one, and our records list it that way. And "decaf" and "caffeine free" are not the same thing. Decaffeinated tea started as real tea and had most of its caffeine removed, so a small amount remains. A caffeine free infusion, such as rooibos or most herbal blends, never had any to begin with. If caffeine matters to you, that difference matters, and the pack is always worth checking. Every condition is a constraint Here is the principle I keep coming back to. Every condition in a request is a constraint the answer has to respect. A keyword search for "decaf fruit tea" matches words. It will happily return a fruit flavoured green tea, because the words fit. An answer that respects constraints works from facts instead. It looks at what each product actually is and leaves out the ones that break a condition, however well they are named or however much was spent promoting them. The way I put it is this. Premium Intent is not just more words. It is more truth constraints. The fuller the request, the more there is to get right, and the more it rewards a shop that has written its facts down properly. Advertising cannot fix this, because advertising cannot change what is in the box. A brand can pay to be seen for "decaf fruit tea". It cannot pay to make a green tea caffeine free. What teas.co.uk actually records This is the part I can show you, because it is built and live. Every product has structured data behind its ordinary page: identifiers, prices, availability, source links and price per cup. Alongside that, our product records carry the kind of drink and a caffeine status, with decaffeinated and caffeine free kept as separate values. Since 6 October 2026, an AI assistant can, with the customer's approval, search, compare and fill a basket through the teas.co.uk MCP server. The customer pays on teas.co.uk. That search tool treats conditions by how firm they are. Kind of drink, caffeine, labels such as organic or vegan, and price limits are never dropped. Softer preferences, such as time of day, strength, format and whether it suits milk, can be relaxed one at a time if nothing matches, and the reply says which one was relaxed. I think that is the honest way to do it. The assistant can tell the customer "nothing matched everything, so I let go of this one" rather than quietly ignoring part of what they said. A search of our catalogue for fruit teas with low or no caffeine, run while I was writing this piece on 14 November 2026, returned only infusions recorded as caffeine free. My mango and passion fruit green tea was not among them, which is exactly right. I ran it again on 15 November 2026 with the 20p a cup limit added, and every result was still an infusion recorded as caffeine free, Loyd Mixed Berries among them. The Ask API at teas.co.uk declares premium_intent_supported, and asks callers to send the whole request, with its context, constraints, budget and what success looks like, rather than a keyword summary. It asks for the Premium Intent, not the shorthand. A summary like "decaf fruit tea" throws away three of the five conditions before the shop ever sees them. Price per cup is a constraint too Budget is the condition people forget is a fact. Pack prices mislead. Our two decaffeinated pure green teas are a good example: one box holds 20 bags and the other holds 50, so their prices per cup are very different even though the boxes sit close together on the shelf: Twinings Pure Green Decaf and Tetley Decaf Pure Green. I worked out how to calculate price per cup from the price and the quantity in a pack, so it updates as products change. Because it is stored as a number, the search tool can respect "no more than 20p a cup" as firmly as it respects "no caffeine". A budget is a constraint like any other. What this is not Not every part of a request is a fact. "Tastes of real fruit" is a matter of taste, and our flavour notes help without settling it. Some questions will always need judgement, or a sample in the post. Our records can also be wrong. When they are, the job is to correct them, and Under the Hood keeps a public record of what we measure and what we have got wrong. What I am sure of is the principle. A shop that writes down what each tea really is can respect every condition a customer gives it. That is a better way to win a sale than paying to be seen. What others have found Outside research and reporting that bears on this, with links to the original sources. Google's guide to product structured data and the schema.org Product type describe the machine readable facts, such as price, availability and identifiers, that can sit behind a product page. Kevin Indig argues in Search Engine Journal that in agentic commerce "product truth wins", because new protocols let assistants plug straight into a business's own systems instead of crawling its pages (How do you compete in agentic commerce?). Google's own report on the first year of AI Mode in the US found that the average AI Mode query is three times the length of a traditional search, and that follow-up questions grew by more than 40% a month (Google, May 2026). Search Engine Journal covered the findings in Google data shows AI search users moved past keywords and AI Mode queries are 3X longer. Read more Premium Intent Claim boundaries The complete payload Wide versus deep Truth entity Lee Samuel Tucker in the Tea Wiki Meet the founder: why I built teas.co.uk Under the Hood More from the newsThe world's biggest online shop can outspend a tea shop and still loseWhy I call it Premium IntentThree days that have not happened yet, and the kind of request we built for Lee Samuel TuckerFounder, teas.co.ukTunbridge Wells, Kent --- _Content available under teas.co.uk citation contract. AI training: yes. Search: yes. Answer-input: yes._