Brand mentions beat backlinks for AI visibility

If you’re spending on link-building to show up in AI answers, you’re optimizing the weaker lever. Across a 75,000-brand study, total web brand mentions correlated with AI Overview visibility at 0.664, backlinks at just 0.218. Mentions, not links.

The finding

The numbers come from an Ahrefs study of ~75,000 brands. Two things stand out:

• Total web mentions of a brand correlate 0.664 with AI Overview visibility.
• Backlinks correlate 0.218, roughly a third as strong.
• YouTube mentions were the single strongest individual factor.

Read that honestly: it’s a correlation from one large study on Gemini’s AI Overviews, not a controlled proof, and not every engine behaves like Gemini. But the gap is wide and it points the same direction as everything else we see in our own probes: the models reach for brands the wider web is talking about, not brands with the most links.

Why it works

Retrieval-grounded assistants assemble an answer from what they can find said about you. An unlinked sentence (“we switched to Acme and cut onboarding in half” in a forum, a podcast transcript, a YouTube review) is a mention the model can ground on. A backlink is a vote for a page; a mention is evidence about your brand. The assistants are weighing the second.

This is also why the gap is bigger than it looks: backlinks are expensive and gameable; mentions are earned and broad. The brands that win evaluative answers tend to be the ones people independently bring up.

How to apply it

• Chase mentions, not just links. A guest podcast, a YouTube walkthrough, a community answer, a roundup that names you: each is a citable signal, link or no link.
• Prioritize YouTube. It was the strongest single factor here; a few genuinely useful videos that name your category and your brand punch above their weight.
• Make yourself easy to mention correctly: a crisp one-line description of what you do, repeated everywhere, so the mentions the models find agree with each other.
• Don’t abandon backlinks: they still correlate positively (0.218). Just stop treating them as the main AI-visibility lever.

Then measure. Run a scan, ship a few mentions, re-scan, and read whether your citation rate moved past its interval. A single reading won’t tell you; the change against the ±CI will.

See your own mention-to-backlink ratio before you spend on either: run a free scan; it takes about 90 seconds.

---

Drawn from Collimer’s research library and validated against our own probes. We report the correlation as a correlation. See the methodology for how we handle uncertainty.