The short version: Entity SEO is not about keywords. It is about whether search engines and AI systems know who you are, what you stand for, and whether they trust what independent sources say about you. If they don’t — you disappear. Not just in Google. In ChatGPT. In Perplexity. In every AI system that is now making buying decisions on behalf of your future customers.
Entity SEO — definition
Entity SEO is the practice of establishing a business, person, or organisation as a verifiable entity within the web’s knowledge graph, supported by independent corroboration signals that AI systems can trust when generating recommendations. It is distinct from keyword SEO — which optimises what your content says — because entity SEO optimises what independent sources confirm about you.
You are not your website
Here’s the thing most entity SEO guides miss, because they’re written by SEOs for SEOs.
You are not your website.
Your website is you talking about yourself. Of course it says you’re excellent — you wrote it. Of course it describes your services perfectly — you chose every word. Of course it positions you as the expert in your field — what else would it do?
Search engines — and increasingly AI systems — know this. They’ve known it for years. And they’ve been doing something about it.
When someone asks Google who the best SEO consultant in Hampshire is, or asks ChatGPT to recommend an AEO specialist, those systems don’t just read your service page and take your word for it. They look at everything else that talks about you. Third-party sources. Independent databases. Publications with no financial interest in saying nice things about you. Reviews from clients who chose to write them, not clients you paid to write them.
Think about the last time you hired a solicitor, or a builder, or any professional you were putting serious money into. You didn’t ring them up, ask if they were good, and say “great, that’ll do.” You checked reviews. You asked for references. You looked for independent confirmation that this person was what they said they were.
Google has been doing exactly that since 2012. AI systems are doing it now. And most businesses don’t realise it’s happening — or that they’re failing it.
The Entity Trust Stack
Before the history, the framework. Because once you understand how entity SEO is structured, everything that follows makes sense.
AI visibility is built on three layers of entity trust. Each depends on the previous. You cannot skip to layer three without layers one and two in place.
Layer 1 — Technical Foundation. Before any entity work has value, search engines need to be able to crawl and index your site cleanly. Broken redirects, crawl errors, JavaScript that prevents indexation — any of these mean your entity signals never reach the systems that need to read them. Fix the plumbing before decorating the house.
Layer 2 — Entity Identity. Your website needs to declare, in machine-readable terms, exactly what entity it represents. Schema markup, structured data, consistent naming, Person and Organisation entities that reference each other bidirectionally. This is what tells search systems who you are.
Layer 3 — Corroboration. Independent sources must confirm what layer two declares. This is where most businesses fail. Your website tells search systems who you are. Corroboration is what other sources say about you — editorially, independently, without financial incentive. Review platforms with verified clients. Editorial citations from journalists with no reason to mention you except that you’re relevant.
These three layers map directly onto the 3Cs Framework (Code, Content, Contextual Linking) that I developed in 2010 on foundations taught to me by David, an SEO consultant from Petersfield, and have now extended for the AI era. Technical Foundation is Code. Entity Identity is Content applied structurally. Corroboration is Contextual Linking 2.0: the independent, editorially credible signals that tell the system your entity is what it claims to be.
Get all three right and AI systems can name you confidently. Miss any one of them and you stay invisible — used as a source, anonymously, with no recommendation and no credit.
What entities actually are (without the jargon)
An entity is a thing that exists independently enough to have its own identity. A person. A business. A location. A concept. A product. A framework.
The key word is independently. Your website mentions your business constantly. But that’s you saying you exist. An entity — in the way Google and AI systems understand the term — exists when other things confirm it.
I have been building websites since 2005. I have a site called southamptonwebdesigner.co.uk that I built in 2009 for a dog walker in Portsmouth. That site ranks number one for “dog walker Portsmouth.” Not number two. Number one. For seventeen years. Through Panda, Penguin, Hummingbird, RankBrain, BERT, Helpful Content, and every other algorithm update Google has introduced.
It hasn’t survived by gaming anything. It has survived because it is what it says it is. A local dog walker, in Portsmouth, backed by genuine reviews, a consistent name and address, and a site that has never tried to be anything other than exactly what it is. The entity is clear. The signals are consistent. The system trusts it.
That is entity SEO working at its simplest. And it scales all the way up to enterprise-level brands.
The thirteen-year journey that most businesses missed
To understand where we are, you need to understand where we came from.
For years, SEOs talked about Latent Semantic Indexing — LSI — as though it were a semantic search system. It wasn’t, really. LSI was a statistical method for finding word co-occurrence patterns in documents. Words that appear near each other frequently probably relate to each other. It was never about meaning. It was about proximity. And the distinction matters, because it shows just how far the actual shift turned out to be.
The fundamental problem with the string era — matching characters rather than concepts — is best illustrated by a single word: “Apple.” To a computer matching strings, “Apple” is just the letters A-P-P-L-E. It has no way of knowing whether you want a fruit, a tech company, a record label, or a pub. It just looks for the text and returns everything that contains it.
Or take “Taj Mahal.” Four completely different things share that name — a mausoleum in India, a casino in Atlantic City, a local Indian restaurant, a blues musician. A string-matching system sees the same letters and returns all of them. The user has to do the disambiguation themselves.
Search “Taj Mahal height” and a modern system knows you mean the building in India. Search “Taj Mahal tour dates” and it knows you mean the musician. Search “Taj Mahal menu” and it knows you mean the restaurant. Same words. Completely different entities. Resolved correctly from context — because the system understands things, not strings.
A string-matching computer is like a toddler who’s been asked to fetch “the red thing.” The toddler will bring you anything red — a ball, a toy, a piece of plastic — because they are matching the surface attribute. Ask a mechanic for “the red thing” while pointing at an engine and they hand you the oil dipstick. They understand context, purpose, and the relationships between the objects in front of them. That is the difference between strings and things.
In May 2012, Google announced the Knowledge Graph — and with it, the phrase “things, not strings.” The commercial implication is immediate. Search “how tall is the President of France?” and Google gives you Emmanuel Macron’s height without you having typed his name. It has identified the entity — the President of France — connected it to the attribute — height — and returned the answer. That is what entity understanding does at scale.
In August 2013, Hummingbird rebuilt the core search algorithm around that entity understanding. Not a bolt-on update like Panda or Penguin — those were modifications to the existing engine. Hummingbird was, in Google’s own analogy at the time, a new engine rather than a new part: the biggest change to the core algorithm since 2001, built around semantic understanding.
RankBrain followed in 2015 — machine learning applied to queries Google had never seen before. BERT arrived in 2019 — bidirectional language understanding, processing which words in a sentence actually carry the meaning. By 2021, MUM was processing information across multiple formats and languages simultaneously.
Every one of those updates was moving in the same direction. Toward understanding things rather than matching strings. Toward trusting relationships rather than counting keywords. Toward knowing who rather than reading what.
In 2026, the destination has arrived. The search landscape has fragmented — Google’s AI Overviews, ChatGPT, Perplexity, Microsoft Copilot, Gemini — each with hundreds of millions of users, each generating answers and naming providers without anyone clicking a link. Your buyers are still out there. They’ve just moved into AI systems and they’re asking more complex questions than they ever put into a search box. “Find me the best SEO consultant for a healthcare technology company in the UK, with proven results and transparent pricing.” That query is not going into Google. And the answer is not ten blue links.
Businesses that still think in keywords are navigating with a 2011 map.
The retrieval surfaces problem — and why it changes everything
Here is something most entity SEO guides do not explain, because it requires understanding how AI systems actually work under the bonnet.
You do not optimise for AI platforms. You optimise for the indexes they query.
Microsoft Copilot is grounded in Bing web search. ChatGPT Search has drawn on Bing’s index alongside OpenAI’s own crawler, OAI-SearchBot, and OpenAI now describes it as using multiple search and content sources. Perplexity runs its own crawler and index. Google AI Overviews and AI Mode retrieve from Google’s index. There is no single AI index. It matters because Bing indexation, which most SEOs treated as an afterthought for a decade, is directly relevant wherever Microsoft infrastructure is involved, and still contributes to ChatGPT Search.
A business that is perfectly indexed in Google but invisible to Bing is invisible to Copilot and to part of what ChatGPT Search can see. The consultant who has comprehensive schema on their WordPress site but never submitted to Bing Webmaster Tools is missing the index that feeds Microsoft’s AI ecosystem. Not a universal prerequisite for every AI answer; an avoidable gap all the same.
Entity work needs to cover all retrieval surfaces: Google Search Console and Bing Webmaster Tools both set up. Google Business Profile and Bing Places both claimed. Schema markup that validates in both ecosystems. A curated llms.txt file that tells the AI crawlers which read it (and Lighthouse’s Agentic Browsing audit, which checks for it) which pages represent your core entity. One boundary to be clear about: Google stated on 15 June 2026 that llms.txt neither helps nor harms visibility in Google Search, including AI Overviews and AI Mode, so it is not a Google signal. The entity does not change. The surfaces it needs to appear on have multiplied.
The three things that determine whether AI systems name you
In March 2026, seostrategy.co.uk ran a diagnostic on its own AI provider visibility. I searched “AEO consultant UK” across Google AI Overview, ChatGPT, and Perplexity. Seven competitors appeared in Google’s AI Overview. A different set appeared in ChatGPT. More than twenty appeared in Perplexity. seostrategy.co.uk appeared on none of them.
Not because the content was wrong. Not because the schema was missing. Not because the technical foundation was poor — seostrategy.co.uk scores 100/100 on PageSpeed Insights, has comprehensive structured data, and has more detailed coverage of AEO and entity SEO than most of the sites that appeared in those answers.
The gap was independent corroboration — layer two of the Entity Trust Stack. The three factors that determine whether AI systems name you specifically are:
1. Your entity must be clearly defined. Consistent naming, consistent schema, a Person entity linked to an Organisation entity, and public identifiers that confirm you are who you say you are.
2. Your entity must be corroborated by independent sources. The Global Legal Entity Identifier Foundation (GLEIF) runs the international financial entity verification system, and it classifies the evidence behind each record in three tiers. We borrowed that taxonomy for a diagnostic because the underlying problem is the same one search and AI systems face: how do you weigh a source that is talking about itself?
The Entity Corroboration Model (Sean Mullins, SEO Strategy Ltd, 2026)
Level 1 — Entity Supplied Only
The only source confirming the entity is the entity itself. Your website. Your blog posts. Your marketing materials. AI systems recognise the content but treat the claims with low confidence — because the source has an obvious incentive to present itself favourably.
What we observe at this level: AI systems use the content as a source. Named recommendations are rare.
Level 2 — Partially Corroborated
Some external sources confirm the entity exists and operates in the claimed space. Review platforms. Business directories. Partner listings. Some independent mention. The entity is beginning to be verifiable but significant gaps remain.
What we observe at this level: occasional citations, and provider recommendations that come and go between runs.
Level 3 — Fully Corroborated
Multiple independent, authoritative sources confirm the entity’s existence, expertise, and claims. Editorial citations from publications with no financial interest in naming you. Verified client reviews on structured platforms. Entity database entries with cross-references. Professional registries.
What we observe at this level: the entity is named as a provider for relevant queries with some consistency. The levels are our diagnostic, not a score any platform publishes.
When an AI answers “who are the best entity SEO consultants in the UK?” the providers it names are, in our testing, almost always ones with a Fully Corroborated evidence trail. Not because they have the best content. Because the independent evidence is strong enough for the system to stake a recommendation on them.
3. Your entity must be associated with your topic in the system’s graph. AI systems do not just ask “does this entity exist?” They ask “is this entity associated with this topic?” Entity graph density matters — the more interconnected the nodes around your entity, the stronger the association signal. Sean Mullins → SEO Strategy Ltd → entity SEO → AI visibility → LLM optimisation → Southampton. Every connection strengthens the confidence that this entity belongs in recommendations for those topics.
The corroboration stack — in priority order
Here is where most entity SEO guides get the ordering wrong. They lead with Wikidata because it is the most technically interesting. Wikidata matters — but editorial sources dominate, and the stack should be understood in the order AI systems weight it.
1. Editorial citations. A journalist naming you as a practitioner in a relevant industry publication has, in every provider-visibility diagnostic we have run, counted for more than anything else on this list. No platform publishes its weighting, so treat the ordering here as our observed priority rather than a disclosed formula. The logic is plain enough: those publications have no financial interest in naming you. That is precisely the signal that makes them credible. The editorial versus advertorial distinction your customers understand instinctively — trusting a newspaper review over a paid advertisement — is the same mechanism AI systems apply to source selection.
2. Structured review platforms with editorial verification. Not Google reviews, which any business can accumulate through social pressure. Clutch verifies clients before publishing reviews and is, in our observed outcomes, the structured review source that most often accompanies a named recommendation for professional services. G2 plays the same role for SaaS. These are the sources AI systems treat as independent confirmation because the platform has done the verification work.
3. Structured entity databases. Wikidata is the machine-readable sister to Wikipedia. Where Wikipedia is for humans, Wikidata is for machines. Google documents Wikidata as one of the sources its Knowledge Graph draws on, and it is one of the few public identity sources that machines read directly rather than infer from prose. An entry for your Person entity, linked to an entry for your Organisation entity, with verified properties, creates the triangulation that allows AI systems to resolve the entity pair with confidence. An incomplete Wikidata entry is worse than no entry — it signals attempted but unverified.
4. Business directories with structured data. Crunchbase is the expected corroboration signal for any professional services entity. Apple Business Connect and Google Business Profile provide consistent NAP (name, address, phone) across local and AI search ecosystems. Companies House data, for UK businesses, is verifiable official record that AI systems can cross-reference.
5. Social and professional graph. LinkedIn carries verified employment history and professional connections — part of the entity graph that connects your person entity to your organisation to your stated expertise.
Entity freshness throughout. This is not a one-time build. AI systems weight recent signals alongside persistent ones. New editorial mentions. New client reviews. Updated Wikidata properties. An entity that was strongly corroborated in 2024 but has had no new signals since can experience what we call citation drift: gradually losing association strength as newer, more actively maintained entities accumulate signals around it. Entity work must be maintained.
What happens when you get this right
Azure Outdoor Living is a bespoke outdoor living company based in Norfolk. Bioclimatic pergolas, glass rooms, aluminium structures. Projects from £20,000 to six figures. Premium clients — hospitality groups, luxury property developers, high-net-worth residential clients.
They scaled to seven-figure annual turnover through organic search. The Lanesborough Hotel, one of London’s most prestigious, found them through Google. Not through a referral. Not through a trade show. Through a search query, at the exact moment a procurement decision was being made.
That is what entity SEO produces at the commercial end. Not traffic statistics. Not keyword rankings. Revenue from the right people at the right moment.
Eco Montessori ranks number one nationally for their core terms. Olliers Solicitors has maintained page-one authority across criminal defence for over a decade. These are not coincidences. They are the result of entities that are consistently built, consistently maintained, and consistently trusted by the systems that mediate discovery.
The entity visibility diagnostic — where are you today?
Run this check now. It takes five minutes.
Open ChatGPT, Perplexity, and Google. Search your service category plus “consultant” or “agency.” If you are a law firm, search “criminal defence solicitors Manchester.” If you are a SaaS company, search “managed file transfer software UK.”
Are you named? If not — and if you have genuine client results and real expertise — apply the AI Provider Visibility Score below. This is a diagnostic framework, not a tool. You can run it manually in under ten minutes.
AI Provider Visibility Score
Score one point for each:
1. Named in Google AI Overview for your primary service category 2. Named in ChatGPT response for your primary service category 3. Named in Perplexity response for your primary service category 4. Named in Copilot response for your primary service category 5. Editorial citation in an industry publication in the last 12 months 6. Clutch or G2 profile with at least three verified reviews 7. Wikidata Person entity with cross-references to Organisation entity 8. Crunchbase entry (Person and Organisation) 9. Google Business Profile with AI-relevant service categories 10. Bing Places claimed and Bing indexed 11. Schema: Person and Organisation bidirectionally linked 12. llms.txt live and curated
Score 0–3: ENTITY_SUPPLIED_ONLY. The corroboration layer is absent. AI systems will use your content but will not name you.
Score 4–7: PARTIALLY_CORROBORATED. Some signals exist. You may appear in AI responses occasionally. Gaps in the corroboration stack are limiting recommendation frequency.
Score 8–12: FULLY_CORROBORATED. Multiple independent sources confirm your entity. AI systems can name you with confidence.
The gap between each tier is almost never more content. It is almost always missing corroboration signals.
Entity SEO in 2030 — and the shift that is still coming
“Entity SEO consultant” shows as sub-10 monthly searches in UK Keyword Planner data as of early 2026. “AEO SEO” is at 170 monthly searches and growing at 767% year-on-year. “AI visibility” is growing at 2,000% year-on-year. “GEO agency” went from near-zero to the fastest-growing agency category in eighteen months.
But the longer-term trajectory goes further than AI search recommendations.
The GLEIF framework was designed for financial entity verification. Banks require it. Regulators mandate it. It exists because commercial systems that transact at scale cannot afford to rely on self-supplied claims. AI agent systems are heading in the same direction. As AI moves from answering questions to making procurement decisions autonomously — Gartner forecasts that at least 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028 — the verification requirements will tighten. Verifiable identity. Machine-readable credentials. Professional registry IDs.
Entity SEO in 2030 will look less like “optimise your schema” and more like “maintain your digital identity stack.” The practitioners who treat their entity as an asset requiring active maintenance — not a technical configuration set up once — will be positioned for a world where the verification bar is higher and the commercial stakes of being named or not named are significantly larger.
The businesses that will own the AI recommendation landscape for entity SEO, AEO, and AI visibility queries in 2028 are building the corroboration infrastructure now. In our experience the entity work required to move from ENTITY_SUPPLIED_ONLY to FULLY_CORROBORATED takes three to six months to compound. We have seen Wikidata changes reflected in Google’s Knowledge Graph within four to eight weeks, though nothing is published to guarantee it. Clutch verification takes days and the first reviews take weeks. Editorial citations take months.
Starting now is not early. It is the correct timing — because the query volume for these terms is forming now, not later. By the time “entity SEO consultant” has meaningful search volume, the positions will already be decided.
Entity corroboration is the layer of this work that determines AI provider visibility specifically — whether AI systems use your content as a source (topical visibility) or name your business as a recommendation (provider visibility). The full framework is at entity corroboration for AI provider visibility.
Entity real estate — the economics of building, acquiring, and transferring entity infrastructure as a digital asset — is at entity real estate: the emerging market for pre-built AI visibility assets.
The universal corroboration stack — the twelve-step implementation guide — is at entity authority: the universal corroboration stack.
If you want to understand where your entity stands today, the AI visibility audit is the starting point.
