Ahrefs analyzed a billion data points across eleven AI search studies, and the clearest signal is that AI visibility tracks with how much people talk about your brand, not how much content you publish.
Ahrefs just published eleven studies analyzing over a billion data points on AI search. There’s a lot in there. YouTube correlations, citation volatility, fabrication rates, the whole landscape mapped in numbers. Most of the coverage will focus on the headline findings. But the number worth sitting with is quieter: 0.194.
That’s the correlation between the number of pages on a site and its AI visibility. Essentially zero. The content volume play that defined the last decade of SEO has almost no measurable relationship with whether AI systems mention your brand.
The volume machine breaks down
For years, the winning strategy was legible and repeatable. Publish more. Cover more keywords. Build topical authority through sheer breadth. The logic was sound in a system where Google crawled pages, indexed them, and ranked them against specific queries. More pages meant more surface area, more ranking opportunities, more traffic.
AI systems don’t work that way. They don’t rank pages. They synthesize answers from patterns learned during training and retrieval. A brand that published 5,000 mediocre articles and a brand that published 200 exceptional ones can look identical to a model that’s reading the internet for context, not crawling it for pages.
The Ahrefs data makes this concrete. Domain Rating correlates weakly with AI visibility (0.266 to 0.326 depending on platform). Content volume is even weaker. The factors that actually correlate are all about how widely a brand gets discussed elsewhere: YouTube mentions (0.737), branded web mentions (0.66 to 0.71), branded anchors.
Not what you publish.
What other people say about you.
YouTube as the surprise signal
The YouTube finding is the one that will get the most attention, and it deserves it. A 0.737 correlation with AI visibility across ChatGPT, Google AI Mode, and AI Overviews is striking. Stronger than Domain Rating. Stronger than backlinks. Stronger than anything else in the study.
The mechanism isn’t mysterious. Both Google and OpenAI have trained their models on YouTube transcripts. OpenAI reportedly used over a million hours of YouTube transcriptions for GPT-4. YouTube isn’t just a platform these models cite; it’s a dataset they learned from. When a brand appears frequently in video titles, descriptions, and transcripts, it gets baked into the model’s understanding of what brands are relevant to what topics.
This isn’t “do YouTube SEO.” It’s closer to: brands that generate enough genuine discussion to show up repeatedly in video content are the same brands that AI systems have deeply internalized. The YouTube mentions are a signal of something broader. Real-world relevance that leaves traces across the web.
What the citation data actually tells you
Several of the findings sound alarming in isolation but tell a more nuanced story together.
AI Overviews change their content 70% of the time between observations, with citations lasting an average of 2.15 days. That sounds like chaos. But semantic similarity stays at 0.95 cosine. The answers are remarkably stable. The sources rotate constantly.
This is a crucial distinction. AI systems aren’t changing their minds about what’s true. They’re swapping out which pages they point to while saying essentially the same thing. For marketers obsessed with “getting cited,” this is uncomfortable. Your page might appear today and vanish tomorrow even though the AI is still making the same point your page supports.
Meanwhile, 67% of ChatGPT’s top 1,000 citations are effectively unreachable for marketers. Wikipedia alone accounts for nearly 30%. Another 24% are homepages. The remaining accessible slice is dominated by “best X” listicles (43.8% of cited page types), with 79% of those updated in 2025 and 76% refreshed within the last 30 days.
The freshness tax
That freshness data deserves its own beat. If you want to be in the small window of pages ChatGPT actually cites, you’re looking at a constant update treadmill. Not “refresh your content quarterly.” More like: if your list post is 31 days old, it’s already falling out of the citation pool.
This is a fundamentally different operating model than traditional SEO, where a strong page could compound traffic for years. AI citations are more like social media impressions. Perishable. Requiring continuous investment to maintain.
And the reward for that investment? Google still sends 345x more traffic than ChatGPT, Gemini, and Perplexity combined. The math on dedicating resources to AI citation optimization doesn’t work for most companies yet.
The real strategic question
The Ahrefs data keeps circling back to the same uncomfortable insight. The things that drive AI visibility (brand mentions, YouTube presence, real-world discussion) are mostly the things you can’t engineer with a content team and an editorial calendar. They’re the byproducts of building a brand that people actually talk about.
Content length doesn’t matter (0.04 correlation). Domain authority barely matters. Page count doesn’t matter. What matters is whether your brand shows up when other people write, film, and discuss topics in your space.
That’s not a tactic. It’s a years-long strategic investment in product quality, thought leadership, community, and distribution that generates genuine, unprompted mentions. The companies already doing this well won’t need to change much. The companies that built their growth on content volume and keyword targeting are staring at a model that doesn’t value those inputs.
Eleven studies. A billion data points. And the clearest takeaway is the oldest advice in marketing: build something worth talking about.
Sources & further reading:
- https://ahrefs.com/blog/ai-overviews-vs-ai-mode/
- https://ahrefs.com/blog/ai-brand-visibility-correlations/
- https://ahrefs.com/blog/ai-vs-made-up-brand-experiment/
- https://ahrefs.com/blog/best-lists-research/
- https://ahrefs.com/blog/short-vs-long-content-in-ai-overviews/
- https://ahrefs.com/blog/b2b-seo-statistics/
- https://ahrefs.com/blog/ai-overview-change/
- https://ahrefs.com/blog/ai-overview-triggers/
- https://ahrefs.com/blog/ai-overviews-international/
- https://ahrefs.com/blog/chatgpts-most-cited-pages/
- https://ahrefs.com/blog/almost-all-clicks-happen-in-the-top-10-results/