Ask ChatGPT: “who are the serious players in my industry?” Three competitors come up. Your brand, though, is nowhere to be found. And yet you outrank them on Google.
This gap is no bug. AI answer engines don’t reward the same signals as a traditional search engine. What matters is being a source that the web treats as trustworthy. And that recognition is built mostly somewhere other than on your own site.
GEO (Generative Engine Optimization) refers to optimizing your visibility within AI-generated answers. Here’s what the data says about the sources models draw from, why third-party media carry so much weight, and what that concretely changes, without falling into the trap of abandoning your site.
Key takeaways
- Among third-party sources, editorial and media content dominates: 20.3% of citations according to the Otterly AI Citation Report (over one million citations analyzed), far ahead of forums (5.9%) and blogs (4.6%).
- Being cited is not the same as ranking well: nearly 90% of the pages cited by ChatGPT sit beyond the 20th organic position.
- The trade press concentrates 69% of chatbots’ media citations, according to the Occurrence/Ifop study.
- The nuance: brand content remains the leading source (52.5%, Otterly). The right strategy combines external proof with genuinely citable owned pages.
What the data says about the sources AI cites
The Otterly AI Citation Report, which analyzes over one million citations across several engines, lays out a clear hierarchy of external sources. At the top of third-party sources: editorial and media content (news sites, online media, publications), with 20.3% of citations. They come well ahead of forums and communities (5.9%), government and institutional sites (4.9%), blogs and independent publishers (4.6%), and encyclopedias such as Wikipedia (3.2%).
The message is clear: when AI cites a source outside a brand, it is first and foremost a media outlet or an editorial publication. Press content is the primary building block of third-party validation, far ahead of any other format. Forums, for their part, carry less weight in overall volume (5.9%), but exceed 10% of citations on ChatGPT and Gemini for product, review, and comparison queries.
A Semrush analysis of 150,000 citations (June 2025) sheds light on a point that seems contradictory: Reddit is the most cited domain there (40.1%), ahead of Wikipedia (26.3%) and YouTube (23.5%). The methodology explains the discrepancy. Reddit is a single, giant domain, whereas the “News / Media” category aggregates thousands of titles. By individual domain, a forum like Reddit stands out; by type of source, editorial and media content win. This same gap between citation and ranking appears in the Semrush study on AI search and SEO traffic.
Takeaway: among third-party sources, editorial and media content lead by a wide margin (20.3%). Forums carry less weight in volume, but overperform on certain commercial queries.
Why AI trusts third parties more
The mechanism is called multi-source corroboration. A forum shows several versions of the same piece of information, along with objections and corrections. An official site shows only one, smoothed over. When a piece of information recurs, phrased differently, from distinct contributors, the model reads it as a signal of reliability.
Be careful not to explain everything through “trust.” Part of Reddit’s weight comes from licensing agreements (around $60M with Google, around $70M with OpenAI) that give models direct access to its data. The truth signal is therefore not the only variable.
The filter remains harsh. According to the teams that track citations, roughly 5% of the pages in a given industry pass the first external authority filter. Below that, a site can be perfectly structured and still remain invisible in the answers.
Takeaway: AI looks for a consensus cross-checked across several sources. A brand present only on its own domain remains an isolated fragment, hard to validate.
Being cited is not necessarily tied to a good ranking
The pages cited by ChatGPT can rank 15th organically in Google’s search results for a single query. In other words, the pool of cited pages and the pool of well-ranked pages overlap through the average ranking across the intermediate queries that AI bots generate (Query Fan Out), but not through the ranking on one or two queries. A page can hold the top position on Google and never be picked up in an AI answer.
Conversely, a lower-ranked page is cited if it’s hosted on a trusted third-party domain and if it partially answers several intermediate questions. That’s also why working on your visibility in AI answers calls for a different logic than positioning alone.
SEO keeps its role. Technically solid sites retain an advantage, and the two disciplines partly overlap. But aiming for a good ranking is no longer enough to be chosen as a source.
Takeaway: Google ranking and AI citation are two sets that overlap only partially. Optimizing one does not necessarily guarantee the other.
The nuance that prevents strategic mistakes
One figure tempers the picture: in that same Otterly report, brand content (published on your own site) remains the leading source, with 52.5% of citations, against 47.5% for all third-party sources combined. Studies diverge on this split (other work places third-party sources in the majority), but they all agree on one point: your site is not out of the game.
The two findings reconcile. External proof gets you into the shortlist: it decides whether the model judges you citable. Your own pages, in turn, supply the factual material the model reuses once you’re in the pool.
External presence is therefore the amplifier, your site the foundation. A brand with a strong external footprint but a poor site will be mentioned without being cited well. A brand with an impeccable site but no external footprint risks never entering the shortlist at all.
Takeaway: external proof opens the door, your structured pages supply the cited content. You have to build both, not choose.
What this concretely changes for your visibility
The priority shifts toward the external footprint, and first toward editorial content. Media and publications form the leading category of cited third-party sources (20.3%), and the trade press concentrates 69% of chatbots’ media citations (Occurrence/Ifop study), far ahead of national press (17%). The sector-specific outlet is no longer a mere awareness lever: it becomes a source that models actively seek out.
Format matters as much as the outlet. In professional services, roughly 81% of cited lists come from neutral third parties, versus 19% from self-promotional lists. An editorial comparison where your brand appears alongside established players carries more weight than content that praises itself alone.
Concretely: target recognized publications and sponsored articles on third-party media in your sector, secure co-citations in comparative pieces, have opinion content signed by identifiable experts, and exist on the platforms models consult (Reddit, YouTube, LinkedIn). Sponsored content platforms like Getfluence provide access to a network of vetted media for this type of placement.
Takeaway: the priority GEO project is no longer just your blog. It’s your presence across recognized external sources, in comparative and neutral formats.
Visibility in AI answers is decided less and less on your own domain and more and more within the ecosystem of sources that models cross-check. Start by auditing your external footprint: across how many recognized outlets is your brand cited, and in what contexts? That is often where, rather than on yet another blog page, the next citations are won. The next step is to measure your share of voice in AI answers, so you can steer instead of guess.
Your questions about AI citations and the weight of third-party media
Why does AI cite forums and the press more than my site?
Because models look for a consensus cross-checked across several independent sources. A forum or a press article brings an outside perspective, with nuances and corrections, whereas your site presents only one smoothed-over version. Information repeated by distinct sources becomes a signal of reliability. Your own messaging, by contrast, is perceived as self-interested.
Do backlinks still matter for being cited by AI?
Yes, but differently. A link from a recognized outlet remains an authority signal, and it also boosts your Google positions. On top of that, models look at unlinked mentions, co-citations, and the consistency of your presence across the web. The link alone is no longer enough: context and source are what count.
Should I stop publishing on my own site?
No. Roughly half of the links cited by ChatGPT still point to company sites. Your site supplies the factual, structured content that AI reuses once it judges you citable. External presence gets you into the shortlist, your own pages supply the material. You need both.
Which sources are cited most by AI?
Across 150,000 citations analyzed by Semrush in June 2025, Reddit came first (40.1%), ahead of Wikipedia (26.3%) and YouTube (23.5%). The hierarchy varies sharply by engine: ChatGPT relies heavily on Wikipedia, Perplexity on Reddit and recent content. No source is universal.
How do I know whether my site appears in AI answers?
Run your strategic queries in ChatGPT, Perplexity, and Google AI Mode, and note whether your brand is cited, with what positioning and against which competitors. Repeat the exercise each week to identify a trend. Citation-tracking and AI share-of-voice tools can then automate the measurement.
