--- type: Article name: "Structured data, explained without the jargon." id: "https://selfe.ai/insights/known-and-matched/structured-data-explained-without-the-jargon" url: "https://selfe.ai/insights/known-and-matched/structured-data-explained-without-the-jargon" publisher: "did:web:selfe.ai" description: "A returns policy that says \"just get in touch if there's a problem\" and one that says \"returns accepted within 30 days, unworn, tags attached\" can describe the same shop. Only one of them can be checked by something that isn't a person using judgement." datePublished: "2026-09-14T15:38:25.120Z" author: The Selfe team --- # Structured data, explained without the jargon. > A returns policy that says "just get in touch if there's a problem" and one that says "returns accepted within 30 days, unworn, tags attached" can describe the same shop. Only one of them can be checked by something that isn't a person using judgement. ## In short - A sentence describes. A fact states what, about what, and where it came from. Most shops only ever write sentences. - Well-written copy and checkable fact are different achievements. A page can have one without the other. - An AI increasingly checks a pre-assembled list of facts, not a product page in real time. Being on that list is a different job from having good copy. A bakery calling its loaf "gluten-free" is making a claim a coeliac customer has to take entirely on trust, unless the claim also says which standard it was tested against and when. An AI agent increasingly isn't reading the bakery's page at all, it's checking a pre-assembled list of exactly this kind of specific, sourced claim, and "gluten-free" with nothing behind it doesn't make that list. ## What makes something a fact rather than a sentence "Gluten-free" is a sentence. "Tested below 20 parts per million against the Codex Alimentarius standard, most recently in March" is a fact. Both might describe the exact same loaf. Only one of them can be checked by something that isn't a person reading carefully and trusting the label, and that's the entire distance between an ordinary product page and what people mean by "structured data." The same split shows up anywhere a shop writes about itself: "just get in touch if there's a problem" versus "returns accepted within 30 days, unworn, tags attached"; "usually quick" versus "dispatched within 1 working day." Neither version of either pair is dishonest. One of them just happens to be readable by something other than a person. A fact has three parts, whether or not anyone writes them out explicitly: what it's about, what's true about it, and where that came from. "Gluten-free" on its own is a sentence, because it's missing the last two parts, tested to what standard, and by whom. "Tested below 20 parts per million against the Codex Alimentarius standard, most recently in March" is a fact: the claim, its exact scope, and the source, all present. Nothing about this requires new software. It requires writing the specific version of the claim instead of the general one, which is usually a smaller job than it sounds, because most shops already know the specific version, they just never wrote it down that way. ## Why plain writing alone doesn't close the gap A beautifully written product description can still fail this test completely. "Baked with care, safe for everyone to enjoy" reads well and answers nothing: not the standard, not the date, not who tested it. A well-edited sentence and a structured, checkable fact are two different achievements, earned in two different ways, and a page can easily have one without the other. Most copywriting makes a sentence read better. Almost none of it makes a sentence checkable, because reading well and being checkable were never the same skill. ## The list an AI agent is checking An AI agent answering a question about a product is increasingly checking a list, assembled in advance, of exactly this kind of fact, what's true, about what, sourced from where, not reading the shop's page in real time the way a browsing customer would. A bakery that's easy to find and pleasant to read but never entered its allergen claim onto that list is invisible to the checking, however good the product photography is. Being on the list is a different job from having good product pages, and it's the one that determines whether an AI can say anything useful, right or wrong, about a product at all. None of this needs a shop to become a data company. It needs the specific version of each claim written down once, clearly, with a source attached, instead of left as a well-meaning general sentence. Selfe turns a shop's real facts, sourced and dated, into the list an AI agent checks against, rather than leaving them as sentences it has to guess at, and keeps that list current. ## Do I need new software to do this? Not to understand it. Most of the work is deciding the specific version of a claim instead of the general one, a writing and record-keeping habit before it's a technical one. ## Is this the same as SEO? No. SEO is about being found by a search engine reading pages. This is about a fact being checkable by something reading a pre-built list, not a page at all. ## Act on this - [Check whether an agent can buy from you](https://selfe.ai/agentic-commerce/buyability-check) — the free readiness scan. - [Discover venues](https://selfe.ai/api/registry/discover) — `POST`, semantic browse across the registry. - [Match a bookable answer](https://selfe.ai/api/registry/match) — `POST` with dates and party size. - [Verify Selfe's identity](https://selfe.ai/.well-known/did.json) — `did:web:selfe.ai`. - [Agent card](https://selfe.ai/.well-known/agent-card.json) — how to connect over A2A or MCP.