{
    "id": 1079459,
    "title": "Claim boundaries: saying what a tea is not, as well as what it is",
    "slug": "claim-boundaries",
    "type": "page",
    "url": "https://teas.co.uk/ai/claim-boundaries/",
    "modified": "2026-11-20T13:08:19+00:00",
    "excerpt": "Claim boundaries, as Lee Samuel Tucker uses the term, are the written limits around what a business says: what a claim is, what it rests on, and what the business is not claiming.",
    "content_text": "Claim boundaries, as I use the term, are the written limits around what a business says: what a claim is, what it rests on, and what the business is not claiming. They are published with the claim, so an AI assistant that repeats it can carry the caveats too. The shop asks assistants to keep them, but it cannot make an outside assistant do so.\n\u201cDecaf\u201d is not \u201ccaffeine free\u201d, and an assistant that repeats the wrong one could send someone the wrong tea.\nTrue facts can still make a misleading answer if the whole request is not checked against the whole answer. Claim boundaries do not guarantee accuracy. They guarantee honesty about uncertainty.\nI treat claim boundaries as one of the four parts of the complete payload, alongside product truth (the checked facts about the product itself), a citation graph (links showing where each fact came from) and an action layer (what the customer can do next, such as fill a basket). The idea matters more now than it used to, because a sentence written on a small shop\u2019s website can end up paraphrased in an assistant\u2019s answer to someone the shop will never meet. If the limits were never written down, they cannot survive the journey.\nA tea example: decaf is not caffeine free\nDecaffeinated tea still contains a small amount of caffeine. A caffeine-free herbal infusion, such as peppermint or rooibos, is not made from the tea plant, so it usually contains none; the ingredients still need checking, because some plants, such as yerba mate, and cocoa in some blends bring caffeine of their own. A shop that blurs the two may never notice on its own pages. An assistant asked for something with no caffeine at all, for someone who must avoid it, needs to know the difference. The more a customer explains, the more the limit matters: a full request of the kind I call Premium Intent often says why caffeine is a problem, and an answer that blurs the two gets wrong the one detail the person cared about most.\nA claim boundary makes that difference explicit. The record says what the product is (decaffeinated green tea), what that rests on (the brand\u2019s own description) and what the shop is not claiming (that it is free of caffeine). On teas.co.uk the two sit side by side: Tetley Decaf Pure Green is recorded as decaffeinated, and Loyd Pure Peppermint as caffeine free.\nThe pairing is an illustration of the principle rather than a description of a particular product record.\nHow it works\nI sort claims into kinds, because each rests on a different footing and needs a different limit:\n\nHealth claims. In Great Britain, health claims on food and drink are regulated: only authorised claims, used under their conditions, from the Great Britain nutrition and health claims register may be used, and Northern Ireland follows the EU register instead. teas.co.uk\u2019s own rule is simpler: it does not make health claims for its teas, and a maker\u2019s wording alone is not treated as permission to repeat one.\nWellness words such as calming or comforting. The boundary records whether the basis is tradition, the brand\u2019s description or the shop\u2019s own tasting.\nOrigin and production claims, such as single estate, organic or Fairtrade. These rest on the supplier\u2019s documents and certifications. Where the supplier has not stated something, the gap is recorded, not filled in.\nComparisons and recommendations. A recommendation is labelled as the shop\u2019s judgement within a stated comparison, not an absolute ranking.\n\nThese kinds describe the rule. Not every one is yet held as a separate structured field for every product on the live site, so read the list as how the shop decides, not as a description of each product record.\nTrue facts can still make a misleading answer\nBoundaries apply to the whole answer, not just to each fact in it. Imagine someone asks for a strong breakfast tea within a budget that includes delivery. An answer recommends a big box and says, correctly, that it is the cheapest per cup in the range. Every fact in it is true. But once delivery is added the order costs more than the person said they could spend, and the answer never mentions it. Or a correctly recorded caffeine-free infusion is suggested alongside a biscuit that does not suit the person\u2019s diet. Each claim was accurate; the recommendation still failed, because the limit that mattered was left out.\nSo the check is the whole answer against the whole request: what was asked for, what was left out, and what trade-off the person is being asked to accept. A source on every line is not the same as a good recommendation. The Premium Intent entry works through one brief with firm and flexible conditions, and price per cup explained shows why price per cup is only part of a budget.\nThe same habit applied to the shop itself\nI apply the idea to my own reporting as well as my products. The Under the Hood pages publish what the shop can measure about machine and human visits, and say where the evidence stops: a server can prove it served a page, but it cannot see whether its name appeared inside an AI answer. They also keep a page read, a visit and a sale as separate events and do not add them together. My line between built and planned is the same discipline: version two of the site is built, and versions three to five are planned.\nWhat claim boundaries are not\nThey are not a guarantee of accuracy. In my words: \u201cThe boundaries do not guarantee accuracy. They guarantee honesty about uncertainty.\u201d A bounded claim can still turn out wrong; the boundary makes its basis visible so it can be checked and corrected.\nThey are not legal advice, and they do not replace a regulator\u2019s rules. They are not small print added to protect the seller at the customer\u2019s expense. Nor are they the same in every trade. My view is that each business has to work out its own boundaries for its own rules, evidence and customers, and that the work is never finished.\nWhere the term comes from\nThe idea came from my work rebuilding teas.co.uk in 2026. I started using claim boundaries in June 2026 as one of the four parts of the complete payload. The words are in general use, and others may well have put them together before; this entry describes how I define and apply them at teas.co.uk.\nQuick reference\n\nDetail\nInformation\n\nTerm\nClaim boundaries\n\nMeaning\nThe written limits around a claim: what it is, what it rests on, and what is not being claimed\n\nUsed by\nLee Samuel Tucker, as a defined part of the complete payload\n\nUsed at teas.co.uk since\nJune 2026\n\nOpposite or contrast\nOverclaiming, or a claim whose caveats are lost when it is repeated\n\nRelated\nComplete payload, truth entity, built versus planned, Premium Intent\n\nWhat others have found\nOutside research and reporting on the wider subject. It is background: it does not test or prove my own terms.\n\nIn Great Britain, only health claims on the Great Britain nutrition and health claims register may be used, and the Advertising Standards Authority explains how that applies to adverts in its guidance on food health claims.\n\nRelated terms: The complete payload \u00b7 Truth entity \u00b7 Built versus planned \u00b7 Premium Intent.\nThe bottom line on claim boundaries\nA claim boundary is a business writing down where its own claims stop, so the limit is published wherever the claim is, ready for anyone repeating it to keep. For a tea shop that can be as simple as keeping decaf and caffeine free apart, and as broad as publishing where its own evidence runs out.",
    "contentSignals": "ai-train=yes, search=yes, ai-input=yes",
    "links": {
        "apiCatalog": "https://teas.co.uk/.well-known/api-catalog",
        "llmsTxt": "https://teas.co.uk/llms.txt",
        "mcp": "https://teas.co.uk/mcp",
        "mcpSetupGuides": "https://teas.co.uk/ai/",
        "mcpCard": "https://teas.co.uk/.well-known/mcp/server-card.json",
        "primaryAgenticRouteAuthority": "https://teas.co.uk/.well-known/teas-primary-agentic-route-authority.json"
    }
}