Schema and entities: making your business quotable by AI
Assistants only quote a business they can identify. Hayley, our senior SEO strategist, on the entity work and structured data that let AI attribute facts to you with confidence, not to a competitor.
Assistants quote entities, not pages
When someone asks an assistant "who does X in Bristol" or "which firm handles Y", the model is not scanning for a keyword and reading back the closest match. It is trying to identify a thing in the world, your business, and decide whether it holds a fact confidently enough to state it. That is the shift entity SEO is built around. Search engines and language models both reason about entities: distinct, disambiguated things with properties and relationships, rather than the raw strings of text on a page. This matters because a business can be mentioned all over the web and still be invisible as an entity. If an assistant cannot tell that the "Bryley" on your about page, the "Bryley Ltd" in a directory and the "Rogue Logic" in a case study are one organisation, it will not risk attributing a fact to any of them. Being quotable starts with being resolvable. The engine has to be able to draw a firm line around you and say, with confidence, this claim belongs to this company.
What structured data actually does
Structured data, usually schema.org markup in JSON-LD, is a machine-readable description of what a page is about, sitting alongside the human-readable content. It does not change what your page says, and it will not lift a weak page up the rankings on its own. What it does is remove ambiguity. Instead of leaving a crawler to infer that a string of digits is a phone number, that a name is your founder, or that a block of text is a service you offer, you state it explicitly in a format built for the purpose. It is worth being honest about the limits, because there is a lot of overclaiming in this space. Schema is not a ranking cheat code and it does not make a claim true. Marking a page up as the wrong type, or asserting a rating the page does not show, earns nothing and can be flagged as a structured data problem. Used properly, though, it is the cleanest way to tell an engine exactly who you are and what you do, so it reads the page the way you intended rather than guessing. That legibility is the groundwork everything else sits on.
Build a graph, not a pile of snippets
The common mistake is treating schema as a checklist: add Organization here, LocalBusiness there, a bit of FAQPage at the bottom, done. Disconnected snippets do far less than they should. The value comes from linking them into a graph, so an engine can resolve your organisation, your people, your services and your locations as one connected identity rather than a scatter of unrelated fragments. In practice that means giving your core entities stable identifiers with @id and referencing them consistently across pages, so your Organization node on the homepage is the same node your service pages and author bios point back to. It means using sameAs to link out to authoritative profiles, your Companies House record, LinkedIn, a Wikipedia entry where one exists, so the engine can cross-check who you are against sources it already trusts. Done well, this is what lets an assistant collapse every mention of you into a single, confident entity, which is the precondition for it quoting you by name.
Consistency is the whole game
Entity clarity is fragile. The fastest way to lose it is contradiction. If your business name, address and founding details differ between your site, your Google Business Profile and the directories you are listed in, you are giving the engine reasons to doubt that these mentions describe the same company, or to merge you with a similarly named one. Assistants are cautious about facts precisely because they are penalised for getting them wrong, so any inconsistency makes them less likely to attribute anything to you at all. The work here is unglamorous and it matters more than the markup itself. It means auditing how you are described everywhere you appear, aligning the details, and making sure the entity you assert in your schema matches the entity the rest of the web sees. A tidy graph on your own site pointing at a messy, contradictory footprint elsewhere will still leave an assistant hedging. Consistency across sources is what turns a plausible identity into a confident one.
Markup has to describe the real page
Structured data is a description, and a description that does not match reality is worse than none. The schema has to reflect what is genuinely on the page and true about the business. If your markup claims a price, a review score or a service you no longer offer, an engine that checks the assertion against the visible page, and they do check, will discount the markup or flag it. Worse, an assistant that surfaces a stale fact about you erodes exactly the trust you were trying to build. This is why schema is not a set-and-forget task. Sites change: services get renamed, people move on, opening hours shift. Markup drifts out of sync with the content it describes and quietly stops working, or starts asserting things that are no longer accurate. Keeping the two aligned, and re-validating when pages change, is part of the ongoing discipline rather than a one-off implementation. The goal is that anything an assistant lifts from your markup is something you would be happy to see quoted back to you.
Knowing whether it is working
You do not have to take any of this on faith. Every type you deploy can be checked against Google's Rich Results Test and the schema.org validator, and confirmed to render for the crawler rather than only appearing after JavaScript runs. That tells you the markup is present, valid and legible. It is the necessary first check, though it is not the finish line. The real question is whether assistants actually cite you and get the facts right, and that is answerable too. You can watch how you are represented in AI answers over time: whether you are named, whether the details are correct, and whether attribution goes to you or to a competitor. When a fact comes back wrong or a rival gets credited for your work, that is usually a signal to trace, an inconsistency in the graph, a contradiction across sources, or markup that has drifted. Treating quotability as something you measure, not something you assume, is what keeps it improving.
Where this sits in the wider picture
Schema and entities are not a separate project bolted onto SEO, and they are not a substitute for genuinely useful, well-written pages. They are the layer that makes good content unambiguous to a machine, so the authority you have earned actually resolves to you. Strong pages that are entity-clear and cleanly marked up are the ones that both rank in ordinary search and get lifted into AI answers, because the same clarity serves both readers. If you are weighing where to start, the honest answer is that it depends on your footprint. Some businesses need the entity graph built and connected before anything else; others have decent markup but a contradictory presence across the web that is quietly costing them attribution. That diagnosis is senior work, which is how we run it here, four practitioners, no juniors handed the parts nobody wants to check. If you want to go deeper on the deliverables behind this article, the links below are the right next step.