Every link-building roadmap in enterprise SEO is still organized around the same handful of numbers: Domain Rating, referring domains, anchor text distribution. Those numbers built entire careers because they predicted rankings. A new research push from Ahrefs says they no longer predict the thing that increasingly matters most, whether a brand shows up when someone asks an AI model a question, nearly as well as a factor most link teams have never tracked: YouTube AI citations, meaning how often a brand gets mentioned in third-party video content that AI systems draw on when assembling an answer.
The correlation that broke the leaderboard
Tim Soulo, Ahrefs' CMO, and Ryan Law, its Director of Content Marketing, spent six months running 14 separate studies against more than a billion data points, then published the results via LinkedIn and X in June 2026. Coverage rippled across the SEO industry within days, for a reason that had nothing to do with novelty for novelty's sake: one number in the data set didn't fit the story link builders have been telling themselves for two decades.
A 0.737 correlation is not a rounding error next to Domain Rating or backlink count. In a data set built from over a billion points, it's the strongest single relationship Ahrefs found between any measurable factor and whether a brand actually surfaces in AI-generated answers. Every other traditional SEO metric in the study, the ones link-building programs have spent years optimizing toward, placed lower. That ordering is the whole story.
| FACTOR TESTED | WHERE IT RANKED | WHAT LINK TEAMS CURRENTLY DO WITH IT |
|---|---|---|
| YouTube mentions | Highest correlation measured (0.737) | Usually nothing; owned by a separate video/social function |
| Domain Rating | Below YouTube mentions | Primary KPI for most link-building roadmaps |
| Referring-domain count | Below YouTube mentions | Standard link-building volume metric |
| Other traditional ranking factors | Below YouTube mentions across the board | Treated as the default proxy for authority |
Why YouTube AI citations outrank Domain Rating
The mechanical explanation is retrieval, not ranking. AI systems answering a question about a product category aren't running the same crawl-and-rank process a search engine runs. They're pulling from a retrieval layer that draws on whatever corroborating evidence exists across the open web, and video transcripts are a large, largely untapped part of that evidence. A backlink tells a crawler one site vouches for another. A YouTube video where a named creator says a brand name out loud, compares it to competitors, and demonstrates the product tells a retrieval system something closer to independent, third-party verification, which is exactly the kind of corroboration the six link types AI models actually trust already puts at the center of a defensible GEO link strategy.
Domain Rating measures link equity accumulated by a domain over time. It says nothing about whether independent third parties are actively talking about a brand right now, in a form a retrieval system can parse into a citable claim. YouTube mentions measure exactly that: current, ongoing, third-party corroboration, delivered in a format, spoken and transcribed language, that maps unusually well onto how large language models ingest evidence. The correlation isn't an accident of the metric. It's a fairly direct read on what these systems are actually built to reward.
It's worth being precise about what the number does and doesn't prove. A 0.737 correlation is not causation, and Ahrefs' own researchers have been careful not to claim that posting more YouTube content mechanically produces more AI citations. Plenty of other explanations could sit underneath the relationship: brands that already have strong organic authority may simply attract more YouTube coverage as a side effect of being well known, rather than the video coverage driving the visibility. That caveat matters and shouldn't get dropped. But even read as pure correlation, a relationship this strong, and this different in kind from anything already sitting in a standard link-building toolkit, is too large to leave unexamined. Teams don't need proof of causation to justify testing a channel; they need a reason to believe the test is worth running, and 0.737 against a billion data points clears that bar easily.
What the study says about who gets cited at all
The YouTube finding sits inside a broader picture from the same research push, and the rest of it explains why third-party corroboration matters so much in the first place. Ahrefs found that 'Best X' listicle pages account for 43.8% of all page types ChatGPT cites, meaning the single most common citation format is exactly the kind of comparative, third-party evaluation content a brand can influence but not directly author on its own site. Our own breakdown of what makes a citation defensible covers the same pattern from the page-content side: comparison and evaluation framing consistently outperforms straightforward product description.
The harder number in the data set is that 67% of ChatGPT's top 1,000 citations come from sources brands can't directly influence at all: Wikipedia at 29.7%, homepages at 23.8%, and app stores at 6.6% make up the bulk of it. Read alongside the YouTube correlation, the pattern is consistent rather than contradictory. AI systems lean hard on independent, third-party corroboration, whether that's an encyclopedia entry, a homepage a competitor can't edit, an app store listing, or a video a brand didn't script. YouTube mentions are simply the corroboration channel a brand has the most realistic ability to actively build, compared to Wikipedia edits or app store rankings, which is exactly why it belongs on a link-building roadmap rather than getting filed away as an interesting but unactionable data point.
Where ChatGPT's top citations come from (source types brands can't directly author)
None of this displaces the core finding. It contextualizes it. Third-party corroboration, generally, is the dominant currency in how these systems decide what to cite. YouTube mentions are simply the specific corroboration type that tested strongest, and the one a link-building program can act on with a repeatable process rather than hoping for an unbiased encyclopedia edit or a favorable app store ranking to materialize on its own.
Building YouTube corroboration into the link program
Treating this as a video-team problem is the mistake most organizations will make with this data, because it's the path of least organizational resistance. A link-building team that reads '0.737 correlation, YouTube' and hands the finding to social media has misread what the number is actually saying. It's a link-building outcome achieved through a different acquisition channel than backlinks, not a separate discipline that happens to share a research report.
The practical version of this looks a lot like existing digital PR and outreach work, aimed at a different destination. Instead of pitching a journalist or blogger for a mention and a link, the target is a YouTube creator with an audience in the relevant category, and the ask is a mention, a comparison slot, or a review, not a hyperlink. The muscle memory transfers. The target list and the pitch don't.
The gap analysis your competitors haven't run yet
Most brands have no idea how they compare to competitors on YouTube corroboration, because nobody on the link-building side has ever been asked to measure it. That's the actual opportunity sitting inside this data, separate from the correlation number itself. A metric this new, this uncorrelated with what teams already track, means most competitive sets are still at zero, or close to it, on deliberate YouTube outreach. The brands that build a process around it in the next two quarters aren't catching up to an established practice. They're establishing one.
The mechanics of the gap analysis are simple and don't require new tooling. Search YouTube directly for the category's core terms and every named competitor, log which channels return results for competitors but not for the brand in question, and rank those channels by relevance and audience fit exactly the way Play 1 above describes. Cross-reference against Ahrefs data where available for channel-level search visibility, since a video that ranks for category search terms carries the corroboration signal into two channels at once, video search and AI retrieval, rather than one.
This matters more in categories where the buying decision genuinely involves watching something work rather than reading about it. Ecommerce and consumer product categories are the clearest case: a shopper deciding between two similar products is unusually likely to watch a comparison video before buying, which means the corroboration and the purchase intent sit in the same moment far more often than they do for, say, enterprise software. Our work with ecommerce clients already treats third-party review and comparison content as core to citation strategy, and the case for extending that into deliberate YouTube outreach is straightforward: it's the same corroboration logic, applied to the channel Ahrefs' data says correlates hardest with actually getting cited. The Johann Wolff case study is a useful reference point for how comparison-driven third-party content compounds inside a single category over time, even when the specific channel changes.
None of this replaces traditional link building. Domain Rating and referring domains still matter for classic search visibility, and nothing in Ahrefs' data argues otherwise. What the 0.737 correlation argues is narrower and more specific: a link-building program that has no answer to the question 'which YouTube creators mention us, and why' is missing the strongest single AI-visibility signal Ahrefs has measured so far, while spending most of its budget on signals that tested weaker. Fixing that doesn't require abandoning the existing roadmap. It requires adding one more line item to it, with an owner, a target list, and a monthly cadence, instead of leaving it as a statistic that got shared once in June and then filed away.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Tyler leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.