AI Search (GEO)
Entity SEO and Knowledge Graph work
Entity SEO is the work of making a business one unambiguous identity everywhere machines look — the same name, address, categories, and facts across the website, the listings, the schema markup, and knowledge bases like Wikidata. AI assistants recommend businesses whose facts agree; disagreement creates doubt, and doubt loses the recommendation.
Every price this business publishes is on one page. See pricing
Last updated
What is an entity, and why do AI systems care?
An entity is a thing a machine can identify without ambiguity — this business, at this address, offering these services. AI systems decide what to recommend by checking whether a business's facts agree across every source they read. Agreement reads as confirmation; contradiction reads as noise, and noisy entities do not get recommended.
The classic example is a business that moves offices: the website shows the new address, Google Business Profile gets updated, and then Foursquare, half the directories and at least one data aggregator go on carrying the old suite number for months. Every assistant reading those sources meets a contradiction — and recommends a competitor whose facts agree.
The order matters. The website is the canonical record everything else is checked against, which is why entity work and web design meet here: a site has to state the business's name, address, services, and service area in visible, server-rendered HTML before consistency work elsewhere can pay off.
| Surface | What it feeds | Typical state for a local business |
|---|---|---|
| Website | Every crawler — the canonical record | Facts drifted from the listings |
| Google Business Profile | Gemini, AI Mode, Maps | Categories and services left half-filled |
| Bing Places | Copilot, and ChatGPT's live lookups | Often filed under a generic or parent-level category |
| Apple Business Connect | Siri and Apple Maps | A category describing a neighbouring trade, not this one |
| Foursquare | ChatGPT's licensed places data | Unreadable from outside — the consumer site closed in 2025 |
| Directory citations | Corroboration for all of the above | Years of accumulated inconsistency |
| Wikidata | Model training corpora | Absent |
Does schema markup get you cited by AI?
Schema markup does not get a business cited by AI — and an honest agency should say so. In a controlled study of 1,885 pages with schema against 4,000 matched controls, Ahrefs measured −4.6% in AI Overviews, +2.4% in AI Mode, and +2.2% in ChatGPT: statistical noise. LLMs read visible HTML, not hidden JSON-LD.
Schema still belongs in the build, for two reasons that are not citations: rich results in classic search, and Knowledge Graph entity reconciliation — a complete Organization record with a sameAs array resolving to every claimed profile is how Google's Knowledge Graph confirms that all those profiles are one business.
The distinction is worth insisting on because the category does not. Vendors sell schema as an AI-citation lever; the controlled evidence says it is not one. We do the markup, and we file it under the honest heading.
What does Knowledge Graph and entity work actually involve?
Knowledge Graph and entity work means making every statement of fact about the business agree, everywhere, and registering the entity where machines look it up. Most of the work is unglamorous reconciliation; the compounding return is that every assistant reading any source gets the same answer about who you are.
- Name, address, phone, hours, and categories made byte-identical across the website, Google Business Profile, Foursquare, Apple Business Connect, Bing Places, and 40+ core directories.
- Organization and LocalBusiness schema with a complete sameAs array that resolves to every claimed profile — the reconciliation signal, verified rather than assumed.
- A Wikidata record — not Wikipedia — because Wikidata sits in essentially every model's training corpus and entity-recognition improvements typically show within 30 to 60 days.
- The website's own pages stating the facts in visible, server-rendered HTML, since AI crawlers execute no JavaScript.
- Ongoing duplicate suppression and consistency defence, because listings drift and aggregators reintroduce old data.
Related: Technical SEO: crawler access and renderingThe anatomy of a citation, and why NAP data drifts
Common questions
- Do I need a Wikipedia page?
- No, and for most local businesses it is not attainable — Wikipedia's notability bar blocks the attempt, and paid editing violates its policies. Wikidata is the realistic target: a structured, open knowledge base that sits in essentially every model's training corpus, where entity-recognition improvements typically show within 30 to 60 days of a clean record landing.
- Is entity SEO just citation building under a new name?
- Citations are one layer of it. Traditional citation work spreads the name, address, and phone number across directories. Entity work adds what that misses: the three listings that actually feed ChatGPT, Siri, and Copilot, schema that reconciles every profile to a single identity, a Wikidata record, and a website that states the facts in visible HTML.
- What does my website have to do with entity SEO?
- The website is the canonical record every other source is checked against. It needs to state the business name, address, services, and service area in visible, server-rendered HTML — AI crawlers execute no JavaScript, so facts rendered client-side effectively do not exist. This is where entity work and web design become the same project.
Find out whether your territory is open
One contract per trade in Clearwater. If yours is open you can execute at the published price today; if a competitor already holds it, it is held until their ninety days are up.
The survey is credited in full against the contract if your territory opens and you take it.