AI systems now answer questions about your business. The problem is that they are often wrong.
Consider the typical situation. A brand’s products, services, expertise, locations, leadership and relationships are spread across dozens of pages. An AI model takes fragments of these pages, assembles them probabilistically, and generates a response. The result is often crazy product namesmade-up leaders, misquoted capabilities, and weak or absent attribution.
This is not a failure of AI models. This is a failure of the medium itself. We built the web around pages, links, and prose. AI retrieval systems need something fundamentally different: a structured layer of meaning and evidence.
The proposal: EntityMap
Feature map has just entered public consultation. It is a new open standard that allows organizations to publish a single structured file. This file declares what the organization knows, maps the relationships between its key entities, and links each assertion to its source evidence.
The consultation runs until June 30, 2026, with an official launch planned for July 1. Over the next 33 days, the project is actively seeking implementation feedback, technical critiques, and real-world testing from developers, SEO professionals, publishers, structured data specialists, and anyone building or relying on AI retrieval systems.
Where EntityMap fits in the standards landscape
EntityMap does not replace existing web standards. It fills a gap that sitemap.xml and schema.org were never designed to fill.
Sitemap.xml tells crawlers what pages exist on a website. Schema.org describes what appears on individual pages. EntityMap tells AI systems what an organization is, what it knows, and how that knowledge connects to the entire website.
This distinction is important. Consider the example of a healthcare organization publishing treatment protocols. With schema.org you can annotate a single page. With EntityMap, you can say the following: “Here are our main treatment areas. Here are the relationships between them. Here is the peer-reviewed evidence supporting each claim. Here is where that evidence is located on our site.” An AI system reading this file obtains a structured view of institutional knowledge rather than reconstructing it from page fragments.
Or consider a SaaS company concerned about how AI systems describe its product. EntityMap allows the company to declare: “We offer feature X. It differs from competitors in Y. Here is the proof: link to documentation, link to case study, link to comparison page. The company no longer needs to rely on an LLM to derive differentiation from scattered web content.
The same logic applies to publishers protecting attribution, law firms clarifying the boundaries of expertise, financial services companies navigating regulatory nuances, and brands concerned about AI misrepresentations.
How EntityMap works
EntityMap is a JSON file published to a predictable location on a domain. It contains three main elements.
Entities the elements covered by the organization are named: products, services, people, concepts, locations, regulations, areas of expertise.
Reports map how these entities connect. Examples: “this product improves this result”, “this person leads this team”, “this regulation governs this service”.
Pieces of evidence support passages from the website, linked to their source URL.
Each piece contains attribution metadata: publisher name, source page, retrieval timestamp. This metadata survives extraction, aggregation, and storage in vector databases. When an AI system generates a response using your content, the chain of evidence remains intact.
The specifications are deliberately minimal. The compliance level includes approximately 12 required fields spread across three objects. Everything else is optional enrichment: custom predicates, cross-fragment resolution, verification status declarations, changelog tracking.
Who should pay attention
If you build Recovery augmented generation (RAG) systemscleaner source data means better chains of reasoning and fewer hallucinations.
If you are an SEO professional, this represents a new AI visibility lever. It works alongside traditional content and link strategies rather than replacing them.
If you are an editor, this is a way of declaring what you know and preserve attribution as your content is disaggregated across AI platforms.
If you are concerned about how AI systems represent your organizationit is a tool for asserting control.
The standard is published under CC BY 4.0. There is no vendor lock-in, no subscription, no proprietary software requirements. Community contribution is open. Source code, specifications and validation tools are all available at GitHub.
What the project expects from you
The consultation period is not ceremonial. The project team actively seeks specific forms of feedback.
Technical feedback on implementation: Have you tried creating an EntityMap for your site or product? What broke? What did you find annoying in practice?
Use case validation: Does this solve a problem you are actually facing? Is something essential missing for your field or industry?
Criticism of the predicate: The standard defines 24 basic predicates (IMPROVES, DEPENDS_ON, MEASURES and others). Are these the right semantic abstractions for your work? Should we add or remove from this list?
Integration Ideas: Are you building a generator? A validator? A dashboard to manage EntityMaps? The project wants to know what tool you are considering.
Industry-specific applications: If you work in healthcare, finance, education, law, or another vertical, what would an EntityMap profile for your industry look like?
The specification is available on Entitymap.org/spec/v1.0. A validator is live on Entitymap.org/validate. The community forum and GitHub repository can be found at github.com/entitymap.
Participants are invited to review the specification, test the implementation, raise issues, suggest improvements, and contribute to the discussion by June 30, 2026.
Important Context: It’s Truly Open
This is a proposed standards from the research and AI community. RV Guhaone of the founders of schema.org, reviewed the project and gave it his approval.
The consultation is truly open. The first phase focuses on technical review and early implementation. Wider adoption, sector-specific applications and research into the wider impact of the standard will follow after the consultation closes.
Why this moment is important
If you’ve spent the last few years watching AI systems distort your work, your clients’ work, or your organization’s expertise, now is the time to shape how that changes.
The bar for entry is low. You should review the specification, test it against a real problem that interests you, and tell the project what you found. These comments will inform the standard before it is finalized.
The consultation lasts 33 days. After that, the adoption phase begins.
Disclosure: I’m the CEO of InLinks and Waikay, both of which support the EntityMap standards proposal.
More resources:
Featured image: optimarc/Shutterstock
