Reddit, Wikipedia, Reviews and News: The Sources AI Cites Most

Most of what an AI engine says about your brand comes from other people’s websites. McKinsey estimates that brand-owned sites are only roughly 5 to 10% of the sources AI search references, and that in consumer goods and financial services more than 65% are publishers, user-generated content and affiliate sites. So the practical question is which outside sources the engines lean on, and what a brand can honestly do on each.

The short version: Reddit, Wikipedia, YouTube, review platforms and independent publications show up again and again. The longer version is that each engine has its own habits, and those habits change.

Which sources do the citation studies actually find?

A citation is a source an AI answer links to or names as its basis. Several large studies have counted them. They use different methods, so the figures are not directly comparable, but the pattern is consistent.

Sources: Profound’s citation pattern study, Ahrefs on the most-cited domains in ChatGPT and Pew Research Center.

Two cautions. First, the mix moves. Semrush tracked about 230,000 prompts weekly in 2025 and found Reddit appeared in about 60% of ChatGPT responses in early August, then about 10% by mid-September, while Wikipedia fell from about 55% to under 20%. Second, being cited is not the same as being named. BrightEdge found ChatGPT mentions brands about 3.2 times more often than it cites them, which means the sources shaping an answer are often invisible in it.

Why do the engines favor these sources?

Here we move from documented facts to reasoning. No AI vendor publishes how it weighs sources, so nobody outside can say for certain.

What is documented: Google signed a data deal with Reddit in February 2024, and OpenAI signed its own Reddit content deal in May 2024. Both give structured access to Reddit’s content. Google also states that to be linked in AI Overviews or AI Mode a page must be indexed and eligible for a snippet. Large, frequently crawled sites clear that bar easily, which may be part of the pattern.

A reasonable inference is that these sources answer the kind of questions people ask AI: “what do real users think of X”, “what is X”, “which X should I buy”. Forums hold opinions, encyclopedias hold definitions, review sites hold verdicts and news sites hold events. Your brand needs to be present, accurately, where those questions get answered.

Reddit: what can a brand ethically do?

Reddit is a collection of communities, each with its own rules, and many of them dislike promotion. That is exactly why its content is valued: it reads as unpaid opinion.

What works within the rules:

• Read the rules of each subreddit before posting anything.
• Post from a clearly identified brand or staff account and answer questions where you have real expertise.
• Fix your product when threads reveal a genuine complaint. Future threads will reflect it.
• Give customers a good reason to talk about you, then let them do it in their own words.

What to avoid: fake accounts, paid upvotes, planted “has anyone tried” posts, or paying people to recommend you without disclosure. That is astroturfing. Beyond getting banned, it can fall under fake review and endorsement rules such as the FTC’s Consumer Review Rule, which bans buying fake social media indicators. This is general information, not legal advice. Our comparison of brand mentions vs Reddit and community marketing covers where each fits.

Wikipedia and Wikidata: what are the limits?

Wikipedia is not a marketing channel. Its conflict of interest guideline says people with a connection to a subject should not edit the article directly and should suggest changes on the talk page instead. Anyone paid to edit must publicly disclose who is paying them. An article also needs independent coverage to survive, which a brand cannot manufacture.

Wikidata, a free, structured knowledge base that many systems draw on, is more open. Its notability policy accepts an item that can be described using serious and publicly available references, even without a Wikipedia article. Accurate entries, with sources, help engines recognize your company as a distinct entity. Our guide to making your brand a known entity walks through Wikidata, Organization schema and sameAs.

Review sites: what is allowed?

Reviews feed AI answers about “best” and “is X any good” questions. They are also the most heavily policed source on this list.

The rules are clear. In the US, the FTC’s review rule has been in force since October 21, 2024 and bans fake reviews, paying for positive sentiment and undisclosed insider reviews. Google Maps bans offering incentives for reviews or asking only happy customers, and Google says it blocked or removed more than 292 million policy-violating reviews in 2025.

What you can do: ask every customer for an honest review, make it easy, reply to reviews publicly (including the bad ones), and keep your listings complete and consistent. This is general information, not legal advice.

News sites, YouTube and independent publications

Independent publications are the broadest category: national news, trade press, niche blogs, local sites and topic magazines. Forbes and Consumer Reports appear in the ChatGPT top domains above, and much of the long tail is smaller sites. Ahrefs found that by 2026 only 37.9% of AI Overview citations came from Google’s top 10 results for that query, so smaller, relevant pages get cited too.

Ways in include digital PR (pitching journalists with real news or data), contributed expert columns, podcast and YouTube appearances, and in-context brand mentions placed on real publications. YouTube deserves attention on its own: in Ahrefs’ late 2025 update, YouTube mentions showed the strongest correlation with AI visibility, about 0.737. That is correlation, not proof that YouTube causes visibility.

Any paid placement should be labeled as the disclosure rules of each market require, and any link in it marked rel=”sponsored” or rel=”nofollow” under Google’s spam policies. This is general information, not legal advice. Our guide to paid brand mention rules explains why, and the brand mention strategy guide covers the source mix.

What should you do with this?

Do not chase one platform because a study ranked it first this quarter. Aim for an accurate, consistent presence across the source types AI keeps returning to, and measure whether it changes how often you are named. Answers vary from run to run, so measure across many prompts rather than trusting a single screenshot.

What to do next

• Ask ChatGPT, Perplexity and Google your buyers’ top five questions and note which sources each answer cites.
• Fix what you control first: your Wikidata entry, review profiles and facts on your own site.
• Get a free AI visibility check to see which sources currently shape what AI says about you.
• If those sources already lead AI to name you, the next step is being recommended near the top of ChatGPT answers, which The Blue Ocean GPT focuses on.