What Is EntityMap? A New AI Search Standard Explained

A few days after last week's Insights post on entity confusion in AI search, Search Engine Journal covered a new proposed open standard called EntityMap. The timing was good. The post I wrote diagnosed a problem. EntityMap is one proposed solution.

What EntityMap actually is

In plain language, EntityMap is a JSON file you'd publish at the root of your domain (entitymap.json). It declares three things:

1. Your entities. What products, services, people, places, and concepts your business covers.
2. Your relationships. How those entities connect to each other and to broader entities in the world. The spec defines 11 typed relationships (INSTANCE_OF, PART_OF, INCLUDES, DEPENDS_ON, AUTHORED_BY, AFFILIATED_WITH, and a few others).
3. Your publisher identity. Who you are, what your domain is, and which external profiles (LinkedIn, Crunchbase, Wikipedia, your government registrations) represent the same business.

Spec v1.0 was published in June 2026 at entitymap.org. The mandatory core is small (about 12 fields across three object types). There's a reference generator from a tools company called Waikay.

The problem it's trying to solve

This connects directly to a problem I covered last week in When ChatGPT Gets Your Business Wrong: Entity Confusion in AI Search.

When AI systems can't tell two similar businesses apart, they blur them. Reviews from one company show up as another's. Wrong founding year. Services the company doesn't even offer, listed as if they did. The buyer doing pre-call research never knows the AI got it wrong. The vendor whose business was misdescribed never sees the signal.

The problem also hits any brand whose name overlaps with everyday objects or concepts. A B2B technology client I worked with has a single-word brand name whose primary meaning is something other than the company itself. When AI systems are asked about the business by name, they describe the everyday meaning first. The company shows up as a footnote, if at all. A prospect I audited last spring had a two-word brand name where each word carries strong meaning in everyday language. The AI confidently described the everyday concepts and treated the actual company as a side reference. Both are entity confusion. Both stay invisible to the brand unless someone is actively checking what AI systems are saying.

EntityMap's pitch is straightforward: instead of letting AI systems stitch together fragmented signals about your business from across the web, you tell them directly, in a structured format they can consume. Your business asserts what it is, who it includes, and how it connects to other things, all in one file.

That's the theory.

How it would change your stack

If EntityMap gets adopted, here's what you'd add and what you wouldn't.

You'd add: a new entitymap.json file at the root of your domain. Even for complex businesses, the file would likely stay readable. The mandatory core is 12 fields. Optional enrichments (more entities, more relationships, more attribution) extend it without breaking conformance.

You wouldn't need to change: your existing schema.org markup, your sitemap.xml, your llms.txt, your robots.txt. EntityMap is additive.

What AI systems would do with it (in theory): treat your entitymap.json as a canonical source for your entity declarations. Cite your version of who you are instead of guessing from fragmented mentions across forums, directories, and competitor mentions.

The entire pitch in one sentence: give AI systems the authoritative answer to "what is this business" instead of letting them assemble one for you.

Why I'm not recommending implementation yet

Three reasons.

No major AI system has announced consumption. Not ChatGPT. Not Perplexity. Not Google AI. Not Anthropic. Not Gemini. EntityMap is a standard waiting for a consumer.

Schema.org already does most of this. The Organization schema, properly implemented with sameAs declarations pointing to your external profiles, expresses the same publisher identity and entity relationships EntityMap proposes. Whether AI systems would weight a dedicated entitymap.json more heavily than well-structured schema.org markup is an open question, and the practical answer right now is "no, because nobody's reading it yet."

The pattern matches llms.txt. The llms.txt standard was proposed in May 2024 as an AI-specific file for telling crawlers what content to read. Two years later, adoption is mixed. Google has formally said they don't use it. SE Ranking analyzed 300,000 domains and found no measurable effect on citation frequency. EntityMap could follow a different arc. It might not. The honest answer right now is that we don't know.

The four signals I'm watching for

When any of these happens, my recommendation changes. Until then, it stays where it is.

1. Adoption announcement from a major AI system. ChatGPT, Perplexity, Google AI, Anthropic, or Gemini saying they crawl and weight entitymap.json files. Even one would shift the calculus.
2. Schema.org incorporating EntityMap relationships into the core vocabulary. This would signal that the broader structured-data community sees enough value in the proposal to absorb it. Schema.org has done this with similar independent proposals before (GoodRelations being the cleanest example).
3. Major SEO tooling adding EntityMap support. Semrush, Ahrefs, Screaming Frog, or similar tools adding validation and generation for entitymap.json would mean the ecosystem expects adoption.
4. Real client data showing AI citation lift correlated with EntityMap presence. A credible third-party study or a multi-site case study showing measurable AI citation improvement on sites with entitymap.json files versus those without.

Any one of these would be enough to start the conversation. All four would mean the standard is real.

What to do right now instead

If you're a marketing leader reading this and you want to actually move the needle on AI entity clarity for your business, here's what's worth your team's time today.

1. Audit your Organization schema for completeness and canonical placement. Does it declare your name, legalName, url, logo, foundingDate, and description? Most sites have this. Many have it incomplete or outdated. The architectural failure I see most often is sites putting full Organization schema on every page, with no clear declaration of which page is the canonical entity URL. That's the schema cannibalization pattern I covered in Schema Cannibalization: When Your Site Confuses Google About Itself. The fix is to designate one canonical entity URL (usually the homepage) via @id and mainEntityOfPage in your schema graph, with the rest of the site referencing back to it instead of competing with it.
2. Add sameAs declarations connecting your domain to your external profiles. LinkedIn, Crunchbase, Wikipedia (if you're listed), your industry directories, your verified social accounts. This is the single highest-leverage entity-disambiguation move on most sites, and the cheapest.
3. Make sure your NAP is consistent everywhere. Name, Address, Phone Number. The same way, on every directory, every social profile, every listing. Tiny variations confuse AI systems. "Smith & Co." vs "Smith and Company" vs "Smith Co" reads as three different businesses to a machine.
4. Audit your About page, leadership pages, and services pages for entity clarity. The same audit I covered in last week's post. Disambiguating language. Founding year. Industry. Services you offer (and just as important, services you don't).

If any of those four are weak, fix those first. EntityMap won't save a site with broken Organization schema, no matter how thoughtfully the entitymap.json is constructed.

EntityMap is a proposal looking for users. Until at least one major AI system says they're consuming it, implementing it is low-cost optionality, not a citation strategy.