Most cited sources in AI search: what the evidence shows

If you want one honest answer to which sources AI cites most, start with the only independent, methodology-transparent dataset and label everything else as vendor research. The most cited sources in AI search are Wikipedia, YouTube, and Reddit, with LinkedIn close behind across engines. That much is stable. The exact share each one gets is not, because almost every per-domain number comes from a tool vendor, the numbers disagree, and they move week to week. This page is part of our AI search statistics hub, and it leads with the independent evidence, then cordons off the vendor data so you can see which is which.

Why lead with Pew and not a slick vendor chart? Because Pew Research Center does not sell an AI-visibility product, it published its method, and it worked from real browsing data rather than a scraped sample. That makes it the anchor. Every other source below broadly agrees on the direction, that third-party and reference sites dominate, but each measures a different slice, so we flag each one and give the spread instead of a single figure.

What sources does ChatGPT cite most?

ChatGPT cites third-party and reference sources far more than any single dominant domain, and earned media makes up most of what it reads. The clearest number here comes from Muck Rack, whose "What Is AI Reading?" study found earned media accounted for about 84 percent of AI citations, with journalism alone at about 27 percent and paid or advertorial content at about 0.3 percent.

Frame that number carefully before you use it. Muck Rack sells PR and media-monitoring software, and its conclusion, that earned media is what AI trusts, so do PR, markets its own product. So this is a vendor figure, not independent fact. The method also matters: Muck Rack analyzed more than 25 million links across ChatGPT, Claude, and Gemini only, not Google AI Overviews and not Perplexity, across 17 industries. Across its three editions the earned-media share ranged from 82 to 89 percent. Read it as strong direction, that outside sources writing about you beat your own website, and pair it with the caveat every time. A second vendor, Fullintel, reported an even higher earned share, which corroborates the direction but not the exact number.

Which websites do AI engines cite most often?

The websites AI engines cite most often are Reddit, Wikipedia, YouTube, and LinkedIn, but which one leads depends on the engine and the study. The safe, publishable claim is this: those four recur near the top, the citation economy is diffuse so no single domain dominates most cuts, and the rankings shift fast. Beyond that, the per-domain numbers all come from tool vendors and they disagree, so here they are with their methods and flagged as vendor data.

• Profound (vendor, sells AI-visibility tools): analyzed 30 million citations and reported Wikipedia leading for ChatGPT while Reddit led for Google AI Overviews and Perplexity. Its own cuts are internally inconsistent on the exact top-source share, which is a reason to treat the point figures with caution.
• Peec AI (vendor): looked at 30 million sources across ChatGPT, Google AI Mode, Gemini, Perplexity, and AI Overviews, and found Reddit first overall, then YouTube, LinkedIn, Wikipedia, and Forbes. It noted Wikipedia is strong for ChatGPT and Perplexity but weak for Google's own platforms.
• Semrush (vendor): tracked 230,000-plus prompts over 13 weeks and documented extreme volatility. The share of ChatGPT answers citing Reddit swung from roughly 60 percent to roughly 10 percent over about six weeks, and the swing itself is the finding, not either endpoint.
• Ahrefs (vendor): reported that Wikipedia, YouTube, and Reddit dominate across assistants, in line with the consensus above.

The honest read is that the leaderboard is real but the exact shares are not settled. When you see a confident "Wikipedia is X percent of ChatGPT" or "Reddit is Y percent of Perplexity," it is one vendor's snapshot of one slice at one moment. Our companion page, how to get cited by LLMs (the evidence), walks through the fuller research if you want the deep-dive.

What does Google AI Overviews cite most?

Google AI Overviews cites Wikipedia, YouTube, and Reddit most, and it leans on government sources more than regular search does. This is the one place we have independent numbers, because Pew Research Center measured it directly. Those three sources were about 15 percent of AI-Overview citations, close to their 17 percent share of standard-result links, so AI Overviews is not wildly different from normal Google here.

The standout is government content. The .gov share rose to about 6 percent of AI-summary links from about 2 percent of standard results, so AI Overviews over-cites official sources. News sites held steady at about 5 percent in both. Pew also found about 88 percent of AI summaries cited three or more sources and the median summary ran about 67 words, so AI Overviews spreads its citations rather than leaning on one page. One caveat to keep: Google publicly disputed Pew's separate click-loss conclusions, though not this citation breakdown. If AI Overviews is your target, our guide to ranking in Google AI Overviews covers the practical steps.

Why do Reddit and Wikipedia get cited so much?

Reddit and Wikipedia get cited so much partly because their content is licensed straight to the engines, not only crawled. Reddit content is licensed to Google, reported at about 60 million dollars a year, though neither company ever confirmed the figure, so treat it as reported. Reddit is also licensed to OpenAI, announced in 2024 on undisclosed terms. In 2026, Yelp signed a deal to license its reviews and business data to OpenAI as well.

That licensing is a real, documented reason some platforms show up again and again. It means presence on those platforms is not just an SEO nicety, it is presence on a source the engine has paid to read. Wikipedia's pull is different: it is treated as a high-trust reference the models already lean on. The takeaway for your business is that being named on these licensed and reference-grade sources carries weight you cannot replicate on your own domain.

Independent versus vendor: the evidence at a glance

Here is every finding on this page in one table, sorted by whether the study is independent or vendor-run, with the caveat that governs how you should use each one. Read the independent row first and weight it highest.

How to use this for your own visibility

You use this data by putting your business on the sources AI already cites, not by chasing one magic domain. The research points to a short, plain list of moves. None of them require a big budget, and all of them build presence where the engines actually read.

1. Be present where AI already cites. Get your business onto earned media, third-party review platforms and directories, and reference sites of Wikipedia-grade quality. Your own website alone is not where the citations come from.
2. Earn mentions, not only backlinks. Aim for other sites to name your business in their own words, not just link to you. Ahrefs found branded mentions correlate with AI visibility more strongly than backlinks, though it calls this correlation, not proof.
3. Focus on the platforms the engines read and license. Concentrate on Reddit, YouTube, review platforms, and your Google Business Profile. Several of these are licensed to the engines, which is part of why they get cited so often.
4. Track which sources name you, per engine, every month. Ask ChatGPT, Perplexity, and Google AI Mode your buyer's questions, note which sources they cite and whether you appear, and repeat monthly because the leaderboard moves. Our guide to tracking brand mentions in AI shows a free way to do this.

The strategy behind all four is the same one behind getting cited by AI: build presence off your own site, on the sources the engines trust. When you want that done for you, it is the core of generative engine optimization.

Common questions about the most cited sources in AI search