Entity building: Wikipedia, Wikidata and the Knowledge Graph for brands

You probably do not need a Wikipedia page, and you probably cannot have one. Wikipedia only accepts companies that independent, reliable sources have already written about in depth. What you do need is for search engines and AI assistants to recognise your brand as one distinct thing with stable facts attached. That is entity building, and most of it happens away from Wikipedia.

What an entity is

An entity is a thing a search engine can identify without relying on a string of words: a company, a person, a product, a place. Google stores entities and the facts about them in its Knowledge Graph. Google’s March 2025 announcement of AI Mode says it can draw on the Knowledge Graph alongside web content.

A brand is a recognised entity when a system can answer three questions without confusion. What is it? Which one is it, among others with a similar name? And what is true about it: who founded it, where it is based, what it sells?

A simple test: search for your brand name. If Google shows a knowledge panel with the right logo and description, the entity is resolved. If it shows a different company, or a mix of several, you have a disambiguation problem.

The four places an entity lives

Place	Who controls it	What it takes	What it gives you
Your own site and profiles	You	The same name, description and facts everywhere, plus Organization markup	A consistent base that every other source can copy
Wikidata	Volunteer editors	A clearly identifiable entity with serious public references	A structured, machine-readable record
Wikipedia	Volunteer editors	Significant independent coverage in multiple reliable sources	A heavily cited article, written by others
Google knowledge panel	Google	Nothing you can apply for. Panels are generated automatically	A visible summary you can claim and correct

Work from the top of the table down. Each row depends on the one above it being clean.

Wikipedia: what notability means

Wikipedia’s notability guideline for organisations and companies, which we read on 4 October 2026, sets the test as “significant coverage in multiple reliable secondary sources that are independent of the subject”. It adds that “trivial or incidental coverage” is not enough.

The guideline is specific about what does not count. Press releases, press kits and similar PR material are not independent sources. Paid or sponsored articles cannot be used to establish notability. So sponsored posts, advertorials and syndicated releases count for nothing here.

So the honest route to a Wikipedia article is upstream of Wikipedia. Earn in-depth coverage in publications with editorial standards, and a volunteer may one day write the article. That makes digital PR the legitimate way in, and it means a business with no independent coverage should stop thinking about Wikipedia for now.

Conflict of interest and paid editing

Wikipedia’s conflict of interest guideline says such editing is “strongly discouraged”. Anyone paid to edit must disclose who is paying them, on whose behalf the edits are made and any other relevant affiliation. Editors with a conflict are asked to propose changes on the article’s talk page and to send new articles through the Articles for Creation process, where an uninvolved editor reviews them.

Services that sell Wikipedia pages exist. Undisclosed paid editing breaks Wikipedia’s terms. It can end in a public ban and in the domain being added to Wikipedia’s spam list, which is worse than having no article.

What you can do without breaking any rule:

• If an article exists and contains an error, post a correction request on its talk page, state your connection and supply an independent source.
• If you publish original research, it can be cited in existing articles on the topic by editors who find it useful. Links from Wikipedia carry nofollow, so the value is visibility and not link credit.

Wikidata: the lower bar

Wikidata is the structured database that sits beside Wikipedia. Its notability policy, read on 4 October 2026, accepts an item that meets any one of three criteria. The relevant one for most brands is that the item “refers to an instance of a clearly identifiable conceptual or material entity that can be described using serious and publicly available references”.

That is a lower bar than Wikipedia’s. A Wikidata item holds statements such as official name, founding date, country and official website, each with a reference.

Two cautions. Wikidata is still edited by volunteers, and an item without serious references does not meet the policy. And the claim that Wikidata feeds knowledge panels and AI answers directly is widely repeated but thinly evidenced. Treat an item as a clean, public record of your facts, not as a ranking tactic.

Knowledge panels

Google’s help pages say “knowledge panels are automatically generated” and that the information comes from various sources across the web, including data partners and open web sources. You cannot apply for one.

Once a panel exists, the subject or an official representative can claim it, get verified and suggest changes. Anyone else can use the feedback link on the panel. For a local firm, consistent business details and local citations are part of the same job.

Markup on your own site

Organization markup lets you state your name, logo, address and official profiles in a form machines read. Its sameAs property lists the URLs that are also you: your Wikidata item, your Wikipedia article if one exists, LinkedIn, Crunchbase and social profiles.

The purpose is disambiguation. Evidence that it changes rankings or AI answers is thin: the case study most often quoted comes from a schema vendor and reached us second-hand. Google’s documentation says you do not need special files or markup to appear in its AI features. Add the markup because it is cheap and reduces confusion, and expect nothing measurable from it alone.

A sensible order of work

1. Fix one version of the facts. Legal name, trading name, short description, founding date, founders, address. Use it on your site, in every profile and in every press boilerplate.
2. Add Organization markup with sameAs links to profiles you control.
3. Claim the profiles that exist: Google Business Profile, LinkedIn, Crunchbase and the review platforms for your sector.
4. Earn independent coverage. Everything below waits on this step, and it is the same work that builds reputation signals.
5. Check Wikidata. If an item exists, make sure it is accurate and referenced.
6. Claim the knowledge panel when one appears.
7. Leave Wikipedia to others. Revisit it only when several independent, in-depth sources exist.

What the evidence supports

Wikipedia is worth having for a brand that qualifies. In Profound’s analysis of 680 million citations from August 2024 to June 2025, Wikipedia was 7.8% of ChatGPT citations, the largest share of any domain. That is a vendor study, and it measures how often Wikipedia is cited, not what a page does for the brand it describes.

Beyond that, the firm evidence runs out. No public study isolates a Wikidata item, a knowledge panel or sameAs markup as a cause of better rankings or AI answers. The stronger correlate is how often independent sources name you, which brand mentions and citations covers. Entity work makes sure those mentions are credited to the right brand.

Where to go next

To see whether assistants describe your brand correctly, read how to measure AI visibility. To catch new coverage that could serve as a reference, set up brand mention monitoring. Our methodology explains how we grade sources such as vendor case studies.