What is the difference between Wikipedia, Wikidata and the Google Knowledge Graph?
Wikipedia is an encyclopedia of prose, Wikidata is a database of structured facts, and the Knowledge Graph is Google's own database of entities. Nearly every mistake in this area comes from blurring them.
Wikipedia holds human-written articles. A subject needs notability: significant coverage in independent, reliable secondary sources. Most companies do not meet it, by design.
Wikidata holds items, not articles. Each item has an identifier (Q243 is the Eiffel Tower) and statements built from property-value pairs: instance of tower, located in Paris, designed by Gustave Eiffel. It is a sister project run by the same Wikimedia Foundation, with a different purpose and a lower bar. As of 10 November 2026 it holds more than 124 million items.
The Google Knowledge Graph is Google's proprietary database of entities and the relationships between them, announced in 2012 under the idea of "things, not strings". It underlies Knowledge Panels, the box beside search results (see how Wikipedia changes your SERP). You cannot edit it directly.
The relationship is not a pipeline you can switch on. Google states that it builds panels automatically from information gathered across the web, and it does not list the sources behind a specific panel.
How does structured data reach Google and AI models?
Structured data gives machines a stable identifier and facts they can read without parsing sentences. Every Wikidata item has a QID that never changes and is independent of language: Q42 is Douglas Adams whether the interface is English, Japanese or Arabic. Facts hang on it as triples, such as Apple Inc. (Q312), property P159 (headquarters location), the item for Cupertino.
What is documented about the links to Google and to AI:
- Google. Freebase, the structured database Google acquired earlier, was merged into Wikidata in 2014 to 2016 with Google's participation, and Wikidata keeps Google's identifiers (properties P2671 and P646). That shows a lasting link between the two ID systems. It does not show that Knowledge Panels are assembled from Wikidata statements, and Google has not said so.
- Language models. Wikipedia was a notable part of the training mix of earlier models (about 3% of GPT-3's training data, about 4.5% of LLaMA's). The Wikimedia Foundation launched a Wikidata Embedding Project in October 2025 to make the database easier for AI developers, and in January 2026 named Google, Microsoft, Perplexity, Amazon, Meta and Mistral among Wikimedia Enterprise partners.
None of this lets you inject content into ChatGPT or Gemini. What you can do is publish clean, well-sourced, consistent facts that make accurate description more likely. That is the sober meaning of entity-based SEO and of our AI visibility work: source quality and structure, not prompt tricks.
Is Wikidata's notability bar lower than Wikipedia's?
Yes. An item is acceptable if it meets at least one of three criteria in the Wikidata notability policy: it has a sitelink to a Wikimedia project, it refers to a clearly identifiable entity described by serious public sources, or it fills a structural need. Wikipedia asks whether independent sources have written about the subject in depth; Wikidata asks whether the thing can be identified and verified.
So a mid-sized company declined at Wikipedia can sometimes have a legitimate Wikidata item, backed by a corporate registry entry, a regulatory filing or an authority record. Three caveats:
- "Lower" is not "none". Items for subjects with no serious references get deleted.
- Wikidata is a factual database, not promotional space.
- An item alone is a weaker signal than a Wikipedia article. It is a foundation, not a finish line.
Who can edit a Wikidata item about their own company?
Anyone can technically, but the community strongly discourages it. The Wikidata self-promotion essay says entries about your own company are strongly discouraged and get stricter scrutiny, and paid work must be disclosed under the Wikidata paid-contributions policy and the Wikimedia Terms of Use. When WikiBusines works on an item, it is disclosed in our name. The same logic applies to Wikipedia; see our explainer on paid editing and disclosure.
How do you get a Google Knowledge Panel?
You cannot order one. Google documents that panels appear automatically, and that an entity can verify one through its own Google account and an official profile such as YouTube or Search Console. Not every panel can be verified, and an agency cannot verify on your behalf, only be added as a manager. A verified owner can suggest changes, which Google says take a few days to review, and can change the photo to one already available. The description and subtitle cannot be edited directly.
How long a panel takes to appear, if it appears at all, is not documented. Practitioners report weeks to months, and sometimes never. We do not quote a timeline. Our Google Knowledge Panel programme runs 3 to 4 months at €3,900, includes a Wikidata item, and promises no panel.
What are entity SEO basics: schema, sameAs and authority records?
Entity SEO is corroboration: the same facts and identifiers pointing at one entity from several independent places.
schema.orgmarkup (OrganizationorPerson, in JSON-LD) on your own site states what entity the site represents. This is the part you fully control.sameAsis a list of URLs inside that markup pointing to other representations of the same entity: your Wikipedia article, Wikidata item, verified social profiles.- Authority records are identifiers kept by institutions. Wikidata can link out to them:
| Identifier | Kept by | Applies to |
|---|---|---|
| VIAF | Library authority files (OCLC) | People and organisations in library catalogues |
| ISNI | ISO name identification standard | Authors, performers, organisations |
| ORCID | Researcher registry | Academics and researchers |
| LEI | Financial-regulation identifier | Legal entities in financial transactions |
| ROR | Research organisation registry | Universities, institutes, labs |
You do not need all of them. Include those you legitimately qualify for, and keep them consistent so the whole graph agrees with itself.
What mistakes make a Wikidata item useless?
Four recurring ones separate an item that exists from one that works:
- Orphan items. Nothing links to the item and it links to nothing. A graph derives meaning from connections.
- Missing references. "Headquarters: Berlin" cited to a corporate registry is a fact; without a source it is an assertion that may be stripped.
- No sitelink. An item with no link to any Wikipedia edition is thinner. This is also why Wikidata does not substitute for an article when one is achievable.
- Duplicates and ambiguity. Two items for one company, or a founder conflated with a same-named person. Merging and disambiguating is often the most valuable cleanup.
What can a clean Wikidata item do, and what can't it?
It can establish a permanent QID, improve disambiguation from same-named entities, publish accurate structured facts, tie your site's schema to an open hub, and give a later Wikipedia article cleaner scaffolding. Item creation itself is quick; our Wikidata work takes 1 to 2 weeks.
It cannot:
- Guarantee a Knowledge Panel. Google decides, and many good entities never get one.
- Create notability. It records verifiable facts, not importance.
- Control AI output. It improves the odds of accurate description, never what a model says.
- Override Google Business Profile data. Address, hours and reviews come from there.
- Survive if promotional or unreferenced. Editors and bots clean up junk items.
An agency that promises a panel or control over ChatGPT's wording is selling something nobody can deliver.
When is Wikidata the right first step?
It is the right first step when the subject is real and verifiable but not yet Wikipedia-notable. That covers four situations:
- Real but not yet notable. There is a registry entry and some trade coverage, but not the independent in-depth sources an article needs. Wikidata gives a legitimate identity now while you build the source base.
- Entity confusion. Google or an LLM mixes you up with a same-named company; a disambiguated item is often the most direct fix. See when ChatGPT is wrong about your company.
- Groundwork before a bigger push. Schema,
sameAsand a clean item sit under a later article. - Unsure you qualify for Wikipedia. A notability audit answers that. If sources support an article, pursue page creation with Wikidata in the same workstream; if not, Wikidata still moves the entity layer forward.
Trying for Wikipedia first without the sources wastes weeks and risks a deletion that makes the next attempt harder. A Wikidata item costs little and does not hurt the Wikipedia case.
How WikiBusines helps
We build and maintain the entity layer, disclosed in our name and without promised outcomes:
- Wikidata entity, from €680: a referenced, disambiguated item with the identifiers you qualify for, in 1 to 2 weeks.
- Notability audit, from €490, credited toward the page if you order it within one month, to decide whether Wikipedia is realistic. Page creation starts at €2,400 for English, 4 to 5 weeks, with 90 days of monitoring and guarantee after publication.
- AI visibility audit, €490: what assistants currently say about you and which sources they draw on.
To talk through where your entity should start, contact us.
Frequently asked questions
No. Wikipedia is an encyclopedia of articles with a notability bar; Wikidata is a database of structured items with a lower bar. Both are run by the Wikimedia Foundation, but each has its own rules and editors.
Not provably. Wikidata stores Google's entity IDs and Freebase was merged into it, but Google says panels are built automatically from sources across the web and does not publish the sources of a given panel. Treat Wikidata as one clean source, not a switch.
No. Panels are generated automatically. When one exists for you, you can verify it through your Google account and an official profile, then suggest changes. Not every panel can be verified.
You may, but the community strongly discourages self-promotion and checks such items more strictly. Paid work must be disclosed. The item still needs serious references and cannot be promotional.
There is no documented timeline. Practitioners report weeks to months, and some entities never get a panel. Anyone quoting a fixed number of days is guessing.
It can improve the odds of accurate, unambiguous description, but no one controls what a model says and answers vary between runs. We do not promise AI output.
A QID is the permanent identifier of a Wikidata item, such as Q42 for Douglas Adams. It stays the same even if the name changes and is independent of language.
Start with Wikipedia only if independent, in-depth sources already exist. If the subject is real but not yet notable, a Wikidata item is the achievable first step while the source base grows.
Authors and editorial policy
Bohdan DubylovskyiCo-founder and CBO, WikiBusines→
Roman MelnykCo-founder & Chief Editor, WikiBusines→The article was prepared by the WikiBusines co-founders. We sell paid Wikipedia services, so we disclose this interest: paid editing is always declared on Wikipedia in WikiBusines' name, and claims about the rules are backed by links to primary sources.
Editorial policy: payment disclosure, conflicts of interest, sources