How Does an Entity Actually Get Into Google's Knowledge Graph?

Where did the Knowledge Graph come from?

Google announced the Knowledge Graph on 16 May 2012, in a post titled Introducing the Knowledge Graph: things, not strings. It described the system as “an intelligent model—in geek-speak, a ‘graph’—that understands real-world entities and their relationships”, built so that a search could be for “things, people or places” rather than for strings of characters that happen to match.

The scale was stated precisely. At launch the graph “currently contains more than 500 million objects, as well as more than 3.5 billion facts about and relationships between these different objects”. Google also named its inputs in that post: the graph was “rooted in public sources such as Freebase, Wikipedia and the CIA World Factbook”, and then “augmented at a much larger scale” from other material.

That sentence is worth holding onto, because it is the most specific public statement Google has ever made about what feeds the graph, and it was written fourteen years ago. Nothing in Google's current documentation names a source that directly. The 2012 post is frequently quoted by vendors as though it described present-day mechanics; it describes the state of a system at launch.

Two distinctions follow from the announcement itself. The Knowledge Graph is a database of entities and the relationships between them. It is not a ranking system, and it is not the web index. A knowledge panel is a surface that displays some of what the graph holds. Being in the graph and having a panel are related but not identical, and neither is the same as ranking well.

What happened to Freebase?

Freebase was the crowd-built entity database Google acquired with Metaweb in 2010, and the first source named in the launch post. It no longer exists. Wikidata's project page on the migration records that the closure was announced on 16 December 2014, that Freebase “announced that it was closing and that it wished to give its resources to Wikidata”, and that the platform became read-only on 31 March 2015.

The handover was partial rather than wholesale. The same project page describes an initial import in 2015 that worked through roughly a tenth of the Freebase collection, with over a million items imported in the first phase and a substantial remainder still unprocessed years afterwards. Freebase data did not simply become Wikidata data, and any account that treats the two as interchangeable overstates what happened.

Google's public interface to the graph has been wound down as well. The Knowledge Graph Search API is deprecated. Google states that “we are migrating this API to Cloud Enterprise Knowledge Graph”, directs new users to that product instead, and attaches a blunt stability warning: “This API is not suitable for use as a production-critical service.”

This matters practically. A company that wants to check whether it is “in the Knowledge Graph” is usually querying that deprecated endpoint, often through a third-party tool that does not mention the deprecation. A result from it is a reading of one API's index at one moment, not a statement of record about what Google knows.

What does Google say feeds a knowledge panel?

Google's help documentation on knowledge panels defines them as “information boxes that appear on Google when you search for entities (people, places, organizations, things) that are in the Knowledge Graph”. It says panels are automatically generated.

On sourcing, the documentation is deliberately general. It attributes panel content to various sources across the web, to data partners who provide authoritative information on specific topics, to open web sources, and to verified entities who have suggested edits to facts on their own panels. Those are categories, not names.

What is absent is the important part. The current documentation does not name Wikipedia. It does not name Wikidata. It does not name any registry, directory or publication. The 2012 launch post named three sources; the operational documentation a reader would consult today names none. Any vendor who tells you that a specific input is documented as producing a panel is describing something Google has not published.

There is also no application process. No form, no queue, no submission endpoint, no eligibility checklist for creation. This is the single most important fact for anyone being sold a panel, and it is established by absence rather than by a statement, which is why it is so easy to talk around.

Is there any documented way in?

There is one mechanism Google documents that a site owner directly controls, and it is narrower than it is usually made to sound. Google's Organization structured data documentation says that adding this markup to your home page “can help Google better understand your organization's administrative details and disambiguate your organization in search results”.

It then splits the properties into two groups. Some “are used behind the scenes to disambiguate your organization from other organizations (like iso6523 and naics)”, while others “can influence visual elements in Search results (such as which logo is shown in Search results and your knowledge panel)”. Read carefully, that is a claim about which logo appears in a panel you already have, not about whether a panel appears.

The sameAs property is the one most relevant to identity. Google defines it as “The URL of a page on another website with additional information about your organization, if applicable”, gives a profile page on a social or review site as the example, and notes “You can provide multiple sameAs URLs”. It is the documented join between your own site and other references to the same entity elsewhere.

Every one of these statements sits under the same caveat, which Google repeats across its structured data documentation: “Google does not guarantee that features that consume structured data will show up in search results.” Markup is how you describe yourself unambiguously. It is not a lever that produces a panel, and Google does not claim it is.

What does claiming actually get you?

Google operates a claiming process, and it is routinely mis-sold as a creation process. The claiming documentation opens with a limit: “Not all knowledge panels are claimable as of now.”

Verification works by proving control of an official presence. Google asks you to sign in to one of the official sites or profiles it lists, naming YouTube, Search Console, Twitter and Facebook. The verification is of the representative, not of the facts in the panel.

What verification unlocks is modest. A manager may “suggest changes to the knowledge panel, and add or remove users”. A contributor may only “suggest changes to the knowledge panel”. In both cases the operative verb is suggest. A verified subject does not edit their own panel; they submit suggestions that Google may or may not accept.

The structural point is that the entire claiming flow presupposes a panel that already exists. Nothing in the documentation describes claiming as a way to bring one into being, and Google's own troubleshooting text assumes you are looking at a panel when you begin. If no panel exists, there is nothing to claim, and the documented process simply does not apply to you.

What this means for anyone buying coverage

The honest summary is that knowledge panels are downstream of being a recognisable entity, and Google does not publish the threshold. Coverage may well contribute to that recognisability, because coverage is part of what exists on the open web about a company. Google does not say so, and we are not going to assert a mechanism the documentation does not contain.

That has a direct commercial implication, including for us. A vendor who promises a knowledge panel is promising an outcome they do not control, through a process that is not documented, with no application to submit and no appeal if it does not appear. The promise is unfalsifiable in advance and unattributable afterwards, which is precisely what makes it saleable.

What can actually be done is duller and checkable. Describe the entity consistently everywhere it appears. Publish Organization markup on the home page, including sameAs URLs pointing at profiles you genuinely operate. Keep names, legal names and identifiers consistent across the places that carry them. Claim the panel if one exists and is claimable. None of this is a guarantee, and Google says as much in writing.

Media Morpheus sells media and podcast placement, so treat this section as interested testimony. The documentation does not support selling a knowledge panel as a deliverable, and we would rather publish that than imply a mechanism we cannot evidence.

Methodology and limitations

This report reviews Google's own published documentation as of 23 September 2026: the Knowledge Graph launch announcement of 16 May 2012, the knowledge panel help pages covering definition and claiming, the Organization structured data reference, and the Knowledge Graph Search API page carrying the deprecation notice. It also uses Wikidata's project documentation on the Freebase migration for the dates and scale of that handover. Every URL cited here was retrieved successfully before publication; we do not link pages we could not open. The central limitation is that documentation describes what Google states, not what Google does. Absence of a documented route into the Knowledge Graph is evidence that no route is published, not proof that no mechanism exists internally. We did not attempt to create a panel, and no experiment here would be capable of isolating the cause if one appeared.

Conclusion

Google documents what a knowledge panel is, what broad categories of material inform it, and how an eligible subject can claim one and suggest changes. It does not document how to cause one to exist. The only entry mechanism published for site owners is Organization structured data, which Google frames as helping it disambiguate an organisation, and which it explicitly declines to guarantee will produce any search feature. The sources named at launch in 2012 — Freebase, Wikipedia, the CIA World Factbook — are no longer named in current documentation, and Freebase itself has been closed for a decade. Anyone selling a panel as a deliverable is selling past the evidence.

Frequently Asked Questions

Sources

1. official Google, 16 May 2012. Introducing the Knowledge Graph: things, not strings. Source of the 500 million objects and 3.5 billion facts figures and the statement that the graph is rooted in Freebase, Wikipedia and the CIA World Factbook.
2. official Google Search Help. Knowledge panel definition and sourcing, including automatic generation and the categories of material that inform panels.
3. official Google Search Help. Claiming a knowledge panel: eligibility limits, verification via official profiles, and the manager and contributor permissions.
4. official Google Search Central. Organization structured data, including the sameAs definition, the disambiguation properties, and the no-guarantee caveat.
5. official Google for Developers. Knowledge Graph Search API, carrying the deprecation notice, the migration to Cloud Enterprise Knowledge Graph and the production-critical warning.
6. official Wikidata. WikiProject Freebase, documenting the December 2014 closure announcement, the March 2015 read-only date and the partial import into Wikidata.
7. official Google, 20 May 2020. A reintroduction to our Knowledge Graph and knowledge panels, by Danny Sullivan. Source of the over 500 billion facts about five billion entities figure. Read October 10, 2026.