The short answerOn B2B SaaS buying questions, AI engines did not cite Wikipedia. Across 11,447 ChatGPT, Perplexity and Google AI Overviews answers (22 October to 20 November 2026), Wikipedia earned 0 of 76,340 citations, and none of the 55 agencies and consultants Peec tracks, 53 of them named by the engines, has a Wikipedia article.
You have probably heard the pitch: get your company a Wikipedia page, because ChatGPT trusts Wikipedia more than any other site. There is a real number behind it. Profound found that Wikipedia holds 47.9% of the citations among ChatGPT's ten most-cited sources. What the pitch leaves out is which questions those citations answer.
So we checked the questions that bring a B2B SaaS company revenue: the ones a buyer types when they are choosing a vendor. We read every source the three engines showed for the 99 buying questions LoudFace tracks in Peec AI, over 30 full days. Wikipedia was read four times and cited zero times.
What 11,447 AI answers cited instead of Wikipedia
The engines made 76,340 citations (22,943 on ChatGPT, 18,549 on Perplexity, 34,848 on Google AI Overviews) across 11,447 answers (3,849 on ChatGPT, 3,849 on Perplexity, 3,749 on Google AI Overviews). Here is where the off-site channels landed, per engine.
| Channel | ChatGPT | Perplexity | Google AI Overviews | All three | Share of 76,340 |
|---|---|---|---|---|---|
| Wikipedia and Wikidata | 0 | 0 | 0 | 0 | 0.00% |
| YouTube | 0 | 0 | 2,964 | 2,964 | 3.88% |
| 7 | 206 | 1,785 | 1,998 | 2.62% | |
| 1 | 1 | 1,392 | 1,394 | 1.83% | |
| Review sites (G2, Clutch, Capterra, TrustRadius, Gartner) | 168 | 112 | 46 | 326 | 0.43% |
| Medium | 0 | 0 | 69 | 69 | 0.09% |
| Press release wires | 7 | 19 | 15 | 41 | 0.05% |
Data from the LoudFace Peec AI project: 99 tracked B2B SaaS buying questions, 22 October to 20 November 2026. A citation is a source an engine showed in its answer.
The four Wikipedia reads all came from Google AI Overviews. Two were the articles on ChatGPT and generative engine optimization. The other two were a surname page titled Kalungi and a disambiguation page titled Ten Speed, and neither is about an agency. Each page was read once and none was shown as a source.
Lists did the work. Pages that Peec classes as listicles earned 42,948 of the 76,340 citations, 56.26% of everything, and they led on every engine (12,799 on ChatGPT, 10,767 on Perplexity, 19,382 on Google AI Overviews). Product pages came second with 9,904 (3,149 on ChatGPT, 2,561 on Perplexity, 4,194 on Google AI Overviews).
Google AI Overviews is the social engine. Almost every YouTube, LinkedIn and Reddit citation came from Google AI Overviews, where YouTube alone was 8.51% of its 34,848 citations. ChatGPT cited Reddit once and LinkedIn seven times. We broke each of these channels out on its own in our studies of Reddit, LinkedIn, YouTube and press releases.
ChatGPT leaned on review sites more than the others did. G2, Clutch and the other review sites earned 168 ChatGPT citations, against 46 in Google AI Overviews. A small channel, but it is ChatGPT's largest off-site channel in the table above.
Wikipedia sits at the bottom of that table with a clean zero. If you are a B2B SaaS marketer who has a Wikipedia project on next quarter's plan because of AI search, that is the line to look at.
AI citations by channel, ChatGPT, Perplexity and Google AI Overviews combined (LoudFace Peec AI project, 11,447 answers, 22 October to 20 November 2026).
None of the 55 tracked agencies and consultants has a Wikipedia article
The citation count could hide something: maybe the engines leaned on Wikipedia from memory and just did not show the link. So we checked the other end. Peec tracks 55 agencies and consultants on these questions and the engines named 53 of them at least once. We looked each one up in the English Wikipedia under its name and four or five company-name variants, with redirects followed, then ran a full-text search for the 15 named most often.
None of the 55 has an English Wikipedia article. Every match was an unrelated topic: Omniscience, Graphite, Optimism, Daydreaming.
| Agency | ChatGPT (of 3,849) | Perplexity (of 3,849) | Google AI Overviews (of 3,749) | All answers (of 11,447) | Share | Wikipedia article |
|---|---|---|---|---|---|---|
| Omniscient Digital | 2,326 | 1,379 | 1,365 | 5,070 | 44.3% | None |
| First Page Sage | 1,326 | 488 | 806 | 2,620 | 22.9% | None |
| Siege Media | 1,459 | 600 | 557 | 2,616 | 22.9% | None |
| iPullRank | 1,276 | 507 | 515 | 2,298 | 20.1% | None |
| Skale | 1,020 | 400 | 675 | 2,095 | 18.3% | None |
| Directive Consulting | 939 | 304 | 565 | 1,808 | 15.8% | None |
| SimpleTiger | 635 | 294 | 635 | 1,564 | 13.7% | None |
| Omnius | 641 | 344 | 541 | 1,526 | 13.3% | None |
| Animalz | 862 | 237 | 306 | 1,405 | 12.3% | None |
Answers that named each agency, from the Peec AI brand report for the same window, and the English Wikipedia API, checked 21 November 2026.
Omniscient Digital was named in 2,326 of 3,849 ChatGPT answers. ChatGPT is the engine that, across all topics, cites Wikipedia more than any other site. It recommended a company with no Wikipedia article in six answers out of ten.
That is the finding we would put in front of anyone selling a Wikipedia page as an AI visibility play: on buying questions, a Wikipedia article was not a requirement for being recommended. The engines named agencies from the lists and pages they read for the answer. That matches what we found when we measured our own brand: the engines named us almost only when they had read one of our pages.
Why 47.9% and zero can both be true
The two numbers measure different things, and the gap between them is the useful part.
Profound analysed 680 million citations across ChatGPT, Google AI Overviews and Perplexity from August 2024 to June 2025. It published two metrics. The famous one is share within the top 10: "Within ChatGPT's top 10 most-cited sources, Wikipedia accounts for nearly half (47.9%) of citations among these leading sources." The other is share of all citations, where Wikipedia is ChatGPT's largest single source at 7.8%.
A 5W press release in May 2026 restated the 47.9% as a share of ChatGPT's top-cited factual sources.
The same Profound data also shows the split by engine. In Google AI Overviews, Wikipedia holds 5.7% of the top-10 share, behind Reddit at 21.0%. In Perplexity, Wikipedia does not make the top 10 at all; Reddit holds 46.7%.
Our reading of the gap, and it is an inference: Wikipedia wins questions about facts. Who founded a company, what a term means, when something happened. A buying question asks for a judgement about vendors, and the pages that answer it are lists, reviews and the vendors' own sites. An encyclopedia article describes a company. It does not tell you which one to hire. Our earlier piece on where ChatGPT gets its information found the same pattern on a smaller sample.
Where Wikipedia does matter: what engines and Google know about you
A zero on buying questions does not make Wikipedia irrelevant. It moves the value somewhere else.
Training data. The GPT-3 paper lists English-language Wikipedia as one of the curated datasets in its training mix: 3 billion tokens, a 3% weight, read 3.4 times over a 300-billion-token run. The Wikimedia Foundation says: "Almost all large language models (LLMs) train on Wikipedia datasets, and search engines and social media platforms prioritize its information to respond to questions from their users." What a model believes about your founding year or your category can trace back to Wikipedia even when the answer shows no link.
Google's Knowledge Graph. When Google launched the Knowledge Graph in 2012, it named Wikipedia as one of its public sources, alongside Freebase and the CIA World Factbook. Google's current knowledge panel help page no longer names Wikipedia; it says the information comes from various sources across the web.
Wikipedia and knowledge panels still go together. In our study of 899 software companies, the 252 with an English Wikipedia article got a knowledge panel for their name 49.2% of the time. The 647 without one got a panel 11.3% of the time. For a Knowledge Graph panel specifically, the split was 46.8% against 2.6%. The study is careful about cause: "The data shows an association; notability can drive both the article and the panel."
So the useful model is two separate jobs:
| What you want | Does a Wikipedia article go with it? | What our data shows |
|---|---|---|
| Google shows a knowledge panel for your name | yes, closely | 49.2% with an article vs 11.3% without (899 software companies) |
| An AI model gets your basic facts right | likely, through training data | Wikipedia is in GPT-3's training mix at a 3% weight |
| An AI engine recommends you on a buying question | no sign of it | 0 of 76,340 citations; none of the 55 tracked agencies and consultants has an article |
| A link from Wikipedia passes ranking credit | unlikely | External links carry rel="nofollow", which Wikimedia says search engines such as Google ignore |
Do not expect schema markup to stand in for the article on the panel job. In the same study, "Organization markup with sameAs did not go with more panels": among older records, 43.1% of companies with it got a panel against 40.9% without. The study's call is to "Keep Organization markup for accuracy, not for a panel." Our guide to entity disambiguation covers that accuracy work: a sameAs list, a sourced Wikidata item and one consistent company name.
Knowledge panel rate by English Wikipedia article (LoudFace study of 899 software companies, Google results via DataForSEO, 17 November 2026).
Can a B2B SaaS company get a Wikipedia article?
Most cannot, and Wikipedia says so itself. Its notability guideline for companies reads: "An organization is generally considered notable if it has been the subject of significant coverage in reliable, independent secondary sources." It adds: "Trivial or incidental coverage of a subject by secondary sources is not sufficient to establish notability." And, bluntly: "Bear in mind that only a small percentage of the world's organizations meet the requirements for a Wikipedia article."
The guideline treats your own press releases and marketing as dependent coverage, which cannot establish notability. It asks for attention that "extends well beyond brief mentions and routine announcements". What it wants is independent coverage that treats your company in depth.
If you try anyway, the editing rules apply to you and to any agency you hire:
- You should not write it yourself. The conflict-of-interest guideline: "Do not edit Wikipedia in your own interests, or in the interests of your external relationships. Paid editing must be disclosed."
- A paid editor must say who pays. "Anyone editing for pay must disclose who is paying them, who the client is, and any other relevant affiliation." The Wikimedia Foundation's Terms of Use make the same disclosure binding.
- The link is unlikely to help your rankings. Wikipedia marks external links nofollow, and Wikimedia's own page notes that search engines such as Google ignore nofollow links when they rank pages.
Our advice to a growth-stage SaaS company: if the press coverage to support an article already exists, an independent editor may write one, and companies with an article got knowledge panels far more often in our data. If that coverage does not exist yet, the coverage is the project, and the article comes after it. In our data, neither path showed any link to AI recommendations on buying questions.
What moves AI recommendations on buying questions instead
The engines describe their own selection in general terms. OpenAI: "ChatGPT ranks search results using multiple factors intended to help users find relevant, reliable information. Placement is not guaranteed." Google says that to be eligible as a supporting link in AI Overviews, a page must be indexed and eligible to be shown in Google Search with a snippet, with "no additional technical requirements", and that meeting them does not guarantee the page is shown. Perplexity says each answer carries numbered citations to its sources and does not say how it ranks them.
The data is more specific. Ranked by where the citations went in our 30 days:
- Third-party lists in your category. 56.26% of all citations, and lists led on all three engines: 12,799 citations on ChatGPT, 10,767 on Perplexity and 19,382 on Google AI Overviews. Being an entry on the lists the engines already read is the biggest lever, and it is an off-site job.
- Your own pages. Product pages took 9,904 citations (3,149 on ChatGPT, 2,561 on Perplexity, 4,194 on Google AI Overviews) and homepages 3,990 (2,409 on ChatGPT, 727 on Perplexity, 854 on Google AI Overviews). ChatGPT cited homepages more than the other two engines combined. A page that states plainly what you do and who it is for gets read.
- Google's social sources. YouTube, LinkedIn and Reddit, for Google AI Overviews in particular.
- Review sites, where ChatGPT leans more than the other two.
Wikipedia sits below all of them on these questions. If you want to see where your own brand stands across ChatGPT, Claude, Gemini and Perplexity, you can run your free AI visibility audit.
Method and limits
We read the LoudFace Peec AI project on 21 November 2026, through the Peec REST API, for the 30 full UTC days from 22 October to 20 November 2026. It covers the 99 questions Peec tracks for us, all about choosing an SEO, AEO or AI search agency for a B2B SaaS company, asked daily on ChatGPT, Perplexity and Google AI Overviews. That gave 11,447 answers, 76,340 citations and 144,964 retrievals. Citations ran 22,943 on ChatGPT, 18,549 on Perplexity and 34,848 on Google AI Overviews. Retrievals ran 75,856 on ChatGPT, 43,316 on Perplexity and 25,792 on Google AI Overviews. A page counts as Wikipedia when its address is on wikipedia.org or wikidata.org, in any language. Brand naming comes from Peec's brand report for the 55 brands it tracks in the project. We checked Wikipedia through its public API: a title lookup for every brand under its name and four or five company-name variants, then a full-text search for the 15 most-named. LoudFace is one of the tracked brands; we left ourselves out of the agency table.
The limits matter:
- One kind of question. These are vendor-choice questions in one category. On factual or definitional questions, Wikipedia is cited heavily, as Profound's all-topic data shows.
- Shown sources only. Peec sees the sources an engine displays. It cannot see what a model learned in training, where Wikipedia sits in the mix.
- Tracked brands only. Naming is counted only for the 55 agencies and consultants in the project.
- Article titles only. The Wikipedia check looks for an article about each company. A company mentioned inside another article, such as a founder's biography, would not be found.
The practical call for a B2B SaaS team: put your AI visibility effort into the lists the engines read for buying questions. Treat a Wikipedia article as a knowledge-panel and facts project that starts once independent coverage exists.

