Google's AI Overviews do not cite the web the way search results do. The Machine Relations Index, released on November 3, 2026, tracked which domains actually appear as citations inside AI Overviews, and the leaderboard looks very different from the results SEOs have spent years optimising for. YouTube and Reddit sit at the top, LinkedIn is close behind, and brand pages barely register. For anyone building links or content to be found in AI search, the sources these systems trust most are not the ones most brands have been chasing.
Across 1,274 observed AI Overviews runs, YouTube was cited in 25.20 percent, Reddit in 18.21 percent and LinkedIn in 13.42 percent. Those three platforms alone account for well over half of all citations in the sample, which means the typical AI Overview answer is built on video, community discussion and professional posts rather than on the polished pages brands publish about themselves. It is a striking result because it inverts the usual search hierarchy, where a brand's own site is expected to be the authority on its own topic.
The pattern makes sense once you think about what an AI answer needs. A YouTube video can demonstrate a process step by step, a Reddit thread can show how real people actually solved a problem, and a LinkedIn post can carry a named professional's view. Brand pages, by contrast, are written to sell, and they tend to repeat the same claims every competitor makes. For link builders, the takeaway is that the citation game is now partly a presence game: being quoted in AI answers increasingly means being present where the answers are sourced.
The picture changes when you look beyond Google. TechRadar leads in ChatGPT answers, appearing in 9.05 percent of 8,343 runs, while Reddit leads in Perplexity at 18.27 percent and in Gemini at 7.25 percent. No single domain dominates everywhere, and each engine has its own habits. ChatGPT leans toward established publisher content, Perplexity leans heavily on Reddit, and Gemini sits somewhere in between. A citation strategy built for one engine will not automatically transfer to another.
That fragmentation matters because it kills the idea of a single AI visibility playbook. Optimising for AI Overviews alone leaves ChatGPT and Perplexity untouched, and each of those engines reaches a large audience of its own. The practical response is to stop treating AI search as one channel and start treating it as several, each with its own preferred sources. The common thread is that all of them favour sources other people already trust and link to, which is exactly the kind of source link building has always aimed to create.
If brand pages barely register in AI Overviews, then the old model of publishing a page and waiting for it to rank is no longer enough on its own. The content that gets cited is content that lives where the models look: video walkthroughs, active community threads and expert commentary on professional networks. That does not mean brand sites are worthless, but it does mean their role is changing. They are the destination a citation points to, not necessarily the citation itself.
For link builders, the opportunity is straightforward and familiar. Earned mentions on the platforms AI systems cite, expert quotes that get picked up, and genuinely useful video and discussion content are all linkable assets, and they now carry a second benefit beyond the link itself: a chance to appear inside the answer. The work has not changed as much as the scoreboard has. Build things other people reference, put them where the models can find them, and measure success by citations as well as by rankings.
Sources: Machine Relations Index, November 3, 2026.
Linkbaits helps you create content that earns citations and backlinks. Start with the free tools and build your first linkable asset this week.
Start free trial