Get a free AI visibility report
Third party AI citations are the reason your own website is no longer the main thing shaping what AI says about your brand. Across large studies, the clear majority of the sources AI engines cite are off-domain: reviews, editorial coverage, comparisons, and community threads. This guide explains what the data shows and how to earn those citations.
For two decades, SEO trained marketers to treat their own website as the center of gravity: publish the page, earn the ranking, capture the click. AI search inverts that. When someone asks ChatGPT, Claude, or Google's AI for a recommendation, the model composes an answer from across the web and cites the sources it trusts, and those sources are overwhelmingly not the brand's own domain. A brand can have a polished website and still be shaped, in the answer a buyer actually sees, by a review site or a comparison article it does not control.
The data on this is unusually consistent, which is rare in a field this new. This guide walks through what the studies show about third party AI citations, why AI engines behave this way, and, most importantly, how to earn the off-domain citations that decide whether your brand shows up. It is a shift in where you spend effort, from polishing owned pages to building a corroborated presence across the sources AI reads. For the wider discipline this sits inside, see our guide to answer engine optimization.
The headline finding comes from Muck Rack's ongoing "What Is AI Reading?" study, run by its Generative Pulse team. Analyzing more than 25 million links cited by ChatGPT, Claude, and Gemini across 17 industries, the May 2026 edition found that earned media accounts for 84% of all AI citations, while paid and advertorial content accounts for just 0.3%. The pattern is not a one-off: across three editions of the study going back to July 2025, earned media has ranged from 82% to 89%. In other words, roughly five out of every six sources an AI cites are third-party, and the share barely moves edition to edition.
That stability matters more than the exact percentage. A single study can be an artifact, but a figure that holds across three editions and tens of millions of links describes a structural feature of how AI answers are built, not a passing quirk. The same directional finding shows up in independent work elsewhere: a Search Engine Land analysis with Evertune of roughly 25,000 of the most-cited URLs found that, across nearly 400 million citations, 63% pointed to ranked listicles rather than brand homepages, reinforcing that off-domain, editorial-style sources dominate the citation pool. When multiple independent measurements point the same way, the conclusion is safe to build a strategy on.
A quick definition keeps the rest of this concrete. An owned citation is when AI links to your own domain. A third-party, or earned, citation is when AI names your brand but the source it cites is someone else's property: a review platform, a comparison article, a news story, an academic or government page, an encyclopedic entry, or a community thread. The Muck Rack data folds journalism, which alone makes up around a quarter of all citations, into that earned bucket alongside other independent sources. The practical point is that these are pages you influence rather than publish, which is exactly what makes them harder to win and more valuable when you do.
This behavior is not arbitrary; it follows from how models decide what to trust. Understanding the mechanism tells you where to aim.
A brand describing itself is a weak signal, because every brand claims to be the best. A third party describing a brand is a stronger one, because it carries the implicit weight of an independent judgment. AI models, trained on the structure of the web, have effectively learned that corroboration across independent sources is a proxy for authority. When several review sites, comparison articles, and news pieces say similar things about a brand, that consensus is what the model draws on, not the brand's own marketing copy. This is why earned citations dominate: they are the sources least likely to be self-serving.
Editorial and community sources also tend to be fresher and more structured for extraction than static brand pages. Review platforms update as new reviews arrive, news coverage is timestamped, and comparison articles are built as scannable, answer-first lists, exactly the format models extract cleanly. Brand homepages, by contrast, are often marketing-led, thin on specific claims, and slow to change. The result is that the sources AI reaches for are disproportionately the ones outside a brand's own domain, and that tilt compounds over time as models re-crawl and re-rank.
Knowing that third party AI citations dominate is only useful if you can act on it. The work moves off your own domain and into the ecosystem of sources AI reads. Here is where to focus.
Start by mapping the sources AI engines reference when they name your brand and your category. This tells you what is already working and where competitors are winning citations you are not. Split what you find into owned and earned, then look at the earned side for patterns: a particular review platform, a specific comparison article, or a community that keeps surfacing. Those recurring sources are your highest-leverage targets. Our guide to tracking brand mentions covers how to build this source map systematically.
Review platforms and category comparison pages are among the sources AI engines lean on most for recommendations. Maintain complete, accurate profiles on the platforms relevant to your category, and earn genuine customer reviews on them. Then pursue inclusion in the comparison articles and ranked listicles that already rank for your category, since those ranked lists are a format AI cites heavily. Being accurately present in the sources AI reads is how you enter the answers it composes.
Because journalism makes up a large slice of AI citations, high-authority editorial coverage is one of the most direct levers available. A mention in a trusted publication or an analyst write-up gives AI engines an authoritative, independent source to draw on. Treat public relations and thought leadership not as brand-awareness activities separate from search, but as direct inputs to AI visibility. Strong earned coverage compounds: each placement adds to the corroboration that makes the next citation more likely.
AI engines cite community platforms like Reddit and Quora heavily, because they contain candid, specific, first-hand discussion. Participate authentically where your buyers gather, answer real questions, and let genuine mentions accumulate over time. Manufactured or promotional presence tends to backfire, both with the communities themselves and with the models that have learned to weight authentic discussion. Sustained, honest engagement is slow but durable, and it feeds a source type that AI reaches for constantly. For how these moves fit a full program, see our guide to improve AI search visibility.
The strategic implication is a reallocation of effort. If most of what AI says about you comes from sources you do not own, then pouring all your energy into on-site content while ignoring your off-domain footprint is spending against the smaller share of the citation pool. Owned content still matters, since it is the leg of the stool you fully control and the place AI verifies claims made elsewhere, but it cannot carry the strategy alone. The teams that win AI visibility treat owned, earned, and community presence as one connected system.
That does not mean abandoning your website. It means sequencing the work correctly: keep your owned pages accurate, structured, and crawlable so they support and corroborate, then invest the larger share of new effort into the earned and community sources that make up the bulk of third party AI citations. Measure the split as you go, because the ratio of owned to earned citations in your own results is the clearest signal of whether the reallocation is working. For the tools that track this, see our roundup of AI visibility tools, and for the on-site side, our guide to optimize for AI search.
The most robust figure comes from Muck Rack's "What Is AI Reading?" study, which analyzed more than 25 million links cited by ChatGPT, Claude, and Gemini across 17 industries. It found earned, third-party media accounts for 84% of all AI citations, with paid and advertorial content at just 0.3%. That 84% figure has held remarkably steady, ranging between 82% and 89% across three editions of the study since July 2025. Independent analyses, including work covered by Search Engine Land showing AI's preference for ranked lists and articles over brand homepages, point in the same direction, which is why the finding is considered structural rather than a single-study artifact.
Muck Rack's Generative Pulse team built a large, diverse prompt set spanning many industries and ran it across web-enabled language models, including versions of ChatGPT, Claude, and Gemini. They then systematically analyzed the responses and the links each model cited, classifying every cited source as owned, earned, or paid. The May 2026 edition covered more than 25 million links across 17 industries. Running the study in repeated editions, from July 2025 onward, is what lets the researchers show the pattern is stable over time rather than a snapshot. Because AI systems change as they are retrained, the team frames each edition as a point-in-time reading rather than a permanent constant.
Three numbers capture it. First, earned third-party media accounts for 84% of AI citations in the latest Muck Rack edition, within a stable 82% to 89% range across editions. Second, paid and advertorial content accounts for just 0.3%, which tells you this is an earned-media challenge, not a paid one. Third, journalism alone makes up roughly a quarter of all cited sources, rising sharply for time-sensitive queries. Together these say that the sources shaping AI answers are overwhelmingly independent and editorial, and that buying your way in is not a meaningful option. The lever is earning coverage and mentions across the sources AI already trusts.
No. Your website still matters, but its role has changed. It is the leg of the strategy you fully control and the place AI often verifies claims that third-party sources make about you, so it needs to stay accurate, well-structured, and crawlable. What the data shows is that owned content cannot carry AI visibility alone, because it makes up only a small share of what AI actually cites. The right approach treats your website as the foundation that corroborates your earned presence, while directing the larger share of new effort toward the reviews, editorial coverage, and community mentions that dominate the citation pool.
Prioritize by what AI already cites in your specific category, which you find by auditing your current citations rather than guessing. In general, review and comparison platforms carry heavy weight for product and vendor recommendations, editorial and news coverage carries authority especially for time-sensitive topics, and community platforms like Reddit are cited constantly for candid, first-hand opinion. The exact mix differs by industry and even by market, since AI engines draw on regional sources too. Start with the recurring sources behind your existing mentions and the ones driving competitor citations you lack, then work outward from there rather than spreading effort evenly across every possible source.
It is slower than on-site changes and compounds over time. Technical fixes to your own pages can register within weeks, but earning reviews, editorial coverage, and community presence builds over months. The upside is durability: once a comparison article, review profile, or news mention exists and gets cited, it keeps contributing to the corroboration that makes future citations more likely. Expect early movement within a quarter if you are actively earning coverage and mentions, with the larger gains accumulating over two to three quarters. Continuous measurement matters here, because it lets you see which earned sources are starting to surface in AI answers and double down on what works.
Track the split between owned and earned citations as a core metric, ideally through a platform that monitors continuously across the major AI engines. The ratio itself is diagnostic: if almost all your citations are owned, you are likely under-indexed on the earned sources that dominate the broader citation pool; if you are earning citations across diverse third-party sources, that breadth tends to be more durable. Beyond the ratio, track which specific sources drive your earned citations so you can see the effect of a new review campaign or editorial placement. Because AI answers shift as models update, continuous tracking gives you trustworthy trends rather than a stale one-time snapshot.
Yes, in the details though not in the principle. Every major engine leans heavily on third-party sources, so the core strategy holds across ChatGPT, Google's AI, Claude, Perplexity, and the rest. What differs is which sources each engine favors and how much overlap there is between them, which is often less than teams expect. One engine may weight a particular review platform or publication more heavily than another, and the same query can surface different sources on different platforms. That is why it pays to track citations per platform rather than in aggregate, and to build presence across a diverse set of trusted third-party sources rather than betting everything on one.
.svg)
.svg)
