# Do Wikipedia and Wikidata shape what AI says about you?

> Wikipedia is in the training data of nearly every AI model, and the biggest AI companies pay for a direct feed of it. Here is what that means for a small business, why writing your own article usually backfires, and what to do instead.

Updated October 2, 2026 · AI Syndicate team · https://www.aisyndicate.com/blog/wikipedia-and-ai-answers/

**Quick answer:** Yes. The Wikimedia Foundation says **every large language model trains on Wikipedia**, and Google, Microsoft, Amazon, Meta and Perplexity use its paid data service. But most small businesses **shouldn't write their own article**: Wikipedia needs independent press coverage and **discourages conflict-of-interest editing**. Instead, earn coverage, keep **sourced, disclosed Wikidata facts**, and request fixes on talk pages.

## What are the key facts about Wikipedia, Wikidata and AI?

| Fact | Detail | Source |
| --- | --- | --- |
| Training data | "Every LLM is trained on Wikipedia content," per the Wikimedia Foundation (2023) | Wikimedia Foundation |
| Paid access | Amazon, Google, Microsoft, Meta, Perplexity and Mistral AI are listed as Wikimedia Enterprise users | Wikimedia Enterprise |
| ChatGPT citations | Wikipedia fell from about 55% of responses to under 20% in mid-Sept 2025, still a top-two domain | Semrush (third-party) |
| Google AI Mode | en.wikipedia.org was the #6 cited site, 4.0% share (US, Sept 2026) | Ahrefs (third-party) |
| Knowledge Graph | Google named Wikipedia as a public source when it launched the Knowledge Graph in 2012 | Google |
| Wikidata | A free, structured knowledge base; its data is CC0, so anyone can reuse it | Wikidata |
| Notability | A Wikipedia article needs "significant coverage" in independent, reliable sources | Wikipedia |
| Conflict of interest | People with a COI are "strongly discouraged" from editing those articles directly | Wikipedia |
| Paid editing | Anyone paid to edit must disclose employer, client and affiliation | Wikimedia Terms of Use |

## Why does AI lean on Wikipedia?

Because Wikipedia is one of the biggest free, sourced, regularly updated collections of facts on the internet, and nearly every AI model learned from it.

The Wikimedia Foundation, which runs Wikipedia, put it plainly in 2023: "every LLM is trained on Wikipedia content." It added that Wikipedia is "almost always the largest source of training data" in those sets. [Wikimedia Foundation](https://wikimediafoundation.org/news/2023/07/12/wikipedias-value-in-the-age-of-generative-ai/) In November 2025 it called Wikipedia one of the "highest-quality datasets in the world for training AI." [Wikimedia Foundation](https://wikimediafoundation.org/news/2025/11/10/in-the-ai-era-wikipedia-has-never-been-more-valuable/)

The big AI and search companies also pay for a direct feed. Wikimedia Enterprise sells APIs for Wikipedia and Wikidata, and its site lists Amazon, Google, Microsoft, Meta, Perplexity and Mistral AI under "Trusted by companies that build AI and search." [Wikimedia Enterprise](https://enterprise.wikimedia.com/)

Google's Knowledge Graph, the fact database behind the info boxes in Google results, has roots in Wikipedia too. When Google launched it in 2012, it named Freebase, Wikipedia and the CIA World Factbook as public sources. [Google](https://blog.google/products/search/introducing-knowledge-graph-things-not/) Today Google's help page says only that Knowledge Graph facts come from "a variety of sources," including licensed data and suggestions from content owners. [Google](https://support.google.com/knowledgepanel/answer/9787176?hl=en)

> **What that means for you.** If your business has a Wikipedia article or a Wikidata entry, AI models have probably read it. If it doesn't, AI builds its picture of you from your website, directories, reviews and news coverage instead. Neither is a guarantee of being recommended.

## How often do AI answers cite Wikipedia?

Often, but less than they used to in ChatGPT. No AI company publishes these numbers, so they come from outside studies.

- **ChatGPT:** Semrush tracked 230,000 prompts from July to October 2025. Wikipedia appeared in roughly 55% of ChatGPT responses before mid-September, then under 20% after. It was still one of ChatGPT's two most-cited domains. [Semrush](https://www.semrush.com/blog/most-cited-domains-ai/)
- **Google AI Mode:** Semrush measured Wikipedia at about 3% of responses in the same period. [Semrush](https://www.semrush.com/blog/most-cited-domains-ai/) Ahrefs ranked en.wikipedia.org #6 in the US in September 2026, with a 4.0% share. [Ahrefs](https://ahrefs.com/blog/most-cited-domains-ai-mode/)
- **Perplexity:** around 0.8% of responses, per Semrush. [Semrush](https://www.semrush.com/blog/most-cited-domains-ai/)

> **Citations are not the whole story.** An answer can rely on what a model learned from Wikipedia during training without linking to it. And these shares move: ChatGPT's dropped by more than half in a few weeks. Treat any study as a snapshot.

## What is Wikidata, and why does it matter for my business?

Wikidata is Wikipedia's sister site for facts in a structured form that machines can read. It calls itself a "free, collaborative, multilingual, secondary knowledge base, collecting structured data." [Wikidata](https://www.wikidata.org/wiki/Wikidata:Introduction)

Each thing gets an item with an ID (a Q-number), and each fact is a statement: founded in, headquarters, official website, and so on. Statements can carry references. All of it is released under CC0, which means anyone, including AI and search companies, can reuse it without asking. [Wikidata](https://www.wikidata.org/wiki/Wikidata:Introduction)

That's why it matters for entities. An entity is a specific, identifiable thing, like your company, as opposed to a phrase. When an AI or search engine is trying to work out whether "Summit Dental" in Denver is the same business as "Summit Dental" in Utah, a clean set of facts with an official website attached helps it tell them apart.

Wikidata's bar is lower than Wikipedia's but real. An item must meet one of three tests. The one most businesses rely on is being "a clearly identifiable conceptual or material entity" that "can be described using serious and publicly available references." [Wikidata](https://www.wikidata.org/wiki/Wikidata:Notability) Its help page says most statements "should be verifiable," backed by sources such as newspapers or reference books. [Wikidata](https://www.wikidata.org/wiki/Help:Sources)

## Should I write a Wikipedia article about my own business?

No, for most small businesses. Wikipedia's rules make it hard, and the article usually doesn't survive.

**First, notability.** Wikipedia only covers a topic once it has "significant coverage in reliable sources that are independent of the subject." Your website, ads and press releases don't count. [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Notability) Most local businesses don't have that coverage yet, and no amount of writing can create it.

**Second, conflict of interest.** Wikipedia says editing articles about "your organization, your clients, your employers, or your competitors is strongly discouraged." [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Plain_and_simple_conflict_of_interest_guide) Its guideline says COI editors should propose changes on the talk page and send new articles through Articles for Creation, not publish them directly. [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Conflict_of_interest) Wikipedia also says it is not a place for "advertising, marketing, publicity, or public relations." [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:What_Wikipedia_is_not)

**Third, paid editing has legal rules.** The Wikimedia Terms of Use require anyone paid to contribute to disclose "each and any employer, client, intended beneficiary and affiliation." That's done on a user page, a talk page or in the edit summary. [Wikimedia Foundation](https://foundation.wikimedia.org/wiki/Policy:Terms_of_Use) Wikipedia adds that editors who don't disclose can be blocked. [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Paid-contribution_disclosure) The requirement covers "any contribution to any Wikimedia Projects," so it applies to Wikidata too. [Wikimedia Meta-Wiki](https://meta.wikimedia.org/wiki/Terms_of_use/Paid_contributions_amendment)

> **Be wary of anyone selling a Wikipedia page.** If the subject isn't notable, the article can be deleted. If the writer hides that they were paid, it breaks the Terms of Use, and the cleanup can draw more attention than having no article at all.

## What can I do instead?

Work on the things Wikipedia and AI both look for: independent coverage and clear, sourced facts.

- **Earn real press.** Local news, trade publications and industry awards are the "independent" sources Wikipedia's notability rule asks for. They also help AI directly, because AI engines read news sites. [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Notability)
- **Fix an existing article the right way.** If you already have a Wikipedia article with wrong facts, post on its talk page, say who you are, and use the `{{edit COI}}` template so a volunteer reviews it. Bring an independent source for each fix. [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Conflict_of_interest)
- **Check Wikidata.** Search for your business. If an item exists, its official website, location and founding date should be right and referenced. Anyone editing it for a business they work for, or are paid by, must disclose that. [Wikidata](https://www.wikidata.org/wiki/Help:Sources) [Wikimedia Foundation](https://foundation.wikimedia.org/wiki/Policy:Terms_of_Use)
- **Link your profiles with sameAs.** Google's Organization markup has a sameAs property for "a page on another website with additional information about your organization," like a social or review profile. [Google](https://developers.google.com/search/docs/appearance/structured-data/organization) Google doesn't name Wikidata there. Our advice: if a Wikidata item or Wikipedia article about you exists, list it in sameAs too.

For transparency: AI Syndicate edits Wikidata as a disclosed paid editor, under the account "Aisyndicate." We name the paying brand in edit summaries and don't create new items under that account. Our [Wikidata disclosure](/wikidata-disclosure/) has the details. We don't create Wikipedia articles.

## How do I get Wikipedia and Wikidata working for me honestly?

This is our advice. It is built on the Wikipedia and Wikimedia rules linked above.

1. **Search first.** Look up your business on Wikipedia and Wikidata. Note what exists, what's wrong and what's missing.
2. **Be honest about notability.** Count the independent articles that cover your business in depth. Directory listings and your own press releases don't count. If you have none, a Wikipedia article isn't the next step.
3. **Earn coverage before chasing an article.** Pitch a local reporter a real story, enter industry awards, or get quoted as an expert. That coverage helps AI whether or not Wikipedia ever follows.
4. **Correct errors through the talk page.** Disclose who you are, use `{{edit COI}}`, cite an independent source, and let a volunteer decide.
5. **Keep Wikidata facts sourced and disclosed.** Each fact should have a reference. If you or someone you pay makes the edit, disclose it as the Terms of Use require.
6. **Make your own site match.** Your name, address, phone, founding year and services should read the same on your site as anywhere else. Add Organization schema with sameAs links to your real profiles.
7. **Never hire a hidden writer.** Undisclosed paid editing breaks the Terms of Use and can get accounts blocked and articles deleted.

## What should be on my Wikipedia and Wikidata checklist?

| Check | Done when | Source |
| --- | --- | --- |
| Searched both sites | You know whether a Wikipedia article or Wikidata item exists | Our advice |
| Notability assessed | You've counted in-depth, independent sources about you | Wikipedia |
| No self-written article | You haven't published an article about your own business | Wikipedia |
| Errors raised on talk page | Each fix has a disclosed `{{edit COI}}` request and a source | Wikipedia |
| Wikidata facts referenced | Website, location and founding date each cite a source | Wikidata |
| Paid edits disclosed | Anyone paid to edit names the employer and client | Wikimedia Terms of Use |
| sameAs links added | Your Organization markup lists your real profiles | Google |
| Press coverage growing | You have a plan to earn independent coverage | Our advice |

## How do I check whether AI is using Wikipedia to describe me?

Ask ChatGPT, Google AI Mode and Perplexity "What is [your business]?" and look at the sources each answer lists. If Wikipedia or Wikidata shows up, check that what they say is right. If it doesn't, see which sites the AI uses instead.

Our [Prompt Simulator](/prompt-simulator/) runs one question across 12 engines at once and shows which sources each one cites. If an AI has your facts wrong, [AI is wrong about your business](/blog/ai-wrong-about-your-business/) covers how to correct the record.

## FAQ

### Does ChatGPT use Wikipedia?

Yes. The Wikimedia Foundation says every large language model is trained on Wikipedia content. Semrush measured Wikipedia in about 55% of ChatGPT responses before mid-September 2025 and under 20% after, still one of its two most-cited domains. [Wikimedia Foundation](https://wikimediafoundation.org/news/2023/07/12/wikipedias-value-in-the-age-of-generative-ai/) [Semrush](https://www.semrush.com/blog/most-cited-domains-ai/)

### Do I need a Wikipedia article to show up in AI answers?

No. Most businesses AI recommends don't have one. AI also draws on your website, reviews, directories and news. Wikipedia only covers topics with significant coverage in independent, reliable sources. [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Notability)

### Can I write or edit my company's Wikipedia article?

Wikipedia strongly discourages it. People with a conflict of interest should propose changes on the talk page and send new articles through Articles for Creation. Paid editors must disclose who pays them. [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Conflict_of_interest) [Wikipedia](https://en.wikipedia.org/wiki/Wikipedia:Paid-contribution_disclosure)

### Can my business have a Wikidata item without a Wikipedia article?

Possibly. Wikidata accepts items for clearly identifiable entities that can be described using serious, publicly available references, without needing a Wikipedia page. Statements should be sourced, and paid edits must be disclosed. [Wikidata](https://www.wikidata.org/wiki/Wikidata:Notability) [Wikimedia Foundation](https://foundation.wikimedia.org/wiki/Policy:Terms_of_Use)

### Which AI companies pay for Wikipedia data?

Wikimedia Enterprise sells API access to Wikipedia and Wikidata. Its site lists Amazon, Google, Microsoft, Meta, Perplexity and Mistral AI among companies that use it. [Wikimedia Enterprise](https://enterprise.wikimedia.com/)

## Sources

1. [Wikipedia's value in the age of generative AI](https://wikimediafoundation.org/news/2023/07/12/wikipedias-value-in-the-age-of-generative-ai/) — Wikimedia Foundation. Read Oct 2, 2026.
2. [In the AI era, Wikipedia has never been more valuable](https://wikimediafoundation.org/news/2025/11/10/in-the-ai-era-wikipedia-has-never-been-more-valuable/) — Wikimedia Foundation. Read Oct 2, 2026.
3. [Wikimedia Enterprise](https://enterprise.wikimedia.com/) — Wikimedia Enterprise. Read Oct 2, 2026.
4. [Introducing the Knowledge Graph: things, not strings](https://blog.google/products/search/introducing-knowledge-graph-things-not/) — Google. Read Oct 2, 2026.
5. [How Google's Knowledge Graph works](https://support.google.com/knowledgepanel/answer/9787176?hl=en) — Google. Read Oct 2, 2026.
6. [Wikidata:Introduction](https://www.wikidata.org/wiki/Wikidata:Introduction) — Wikidata. Read Oct 2, 2026.
7. [Wikidata:Notability](https://www.wikidata.org/wiki/Wikidata:Notability) — Wikidata. Read Oct 2, 2026.
8. [Help:Sources](https://www.wikidata.org/wiki/Help:Sources) — Wikidata. Read Oct 2, 2026.
9. [Wikipedia:Notability](https://en.wikipedia.org/wiki/Wikipedia:Notability) — Wikipedia. Read Oct 2, 2026.
10. [Wikipedia:Conflict of interest](https://en.wikipedia.org/wiki/Wikipedia:Conflict_of_interest) — Wikipedia. Read Oct 2, 2026.
11. [Wikipedia:Paid-contribution disclosure](https://en.wikipedia.org/wiki/Wikipedia:Paid-contribution_disclosure) — Wikipedia. Read Oct 2, 2026.
12. [Wikipedia:Plain and simple conflict of interest guide](https://en.wikipedia.org/wiki/Wikipedia:Plain_and_simple_conflict_of_interest_guide) — Wikipedia. Read Oct 2, 2026.
13. [Wikipedia:What Wikipedia is not](https://en.wikipedia.org/wiki/Wikipedia:What_Wikipedia_is_not) — Wikipedia. Read Oct 2, 2026.
14. [Terms of Use](https://foundation.wikimedia.org/wiki/Policy:Terms_of_Use) — Wikimedia Foundation. Read Oct 2, 2026.
15. [Terms of use/Paid contributions amendment](https://meta.wikimedia.org/wiki/Terms_of_use/Paid_contributions_amendment) — Wikimedia Meta-Wiki. Read Oct 2, 2026.
16. [Organization structured data](https://developers.google.com/search/docs/appearance/structured-data/organization) — Google Search Central. Read Oct 2, 2026.
17. [The Most-Cited Domains in AI: A 3-Month Study](https://www.semrush.com/blog/most-cited-domains-ai/) — Semrush. Read Oct 2, 2026.
18. [The 50 Most-Cited Websites in Google AI Mode (September 2026)](https://ahrefs.com/blog/most-cited-domains-ai-mode/) — Ahrefs. Read Oct 2, 2026.