Mentions vs. Links: What Actually Predicts AI Visibility (Data From 75,000 Brands)

The study: 75,000 brands, 11 factors

For years, the most reliable lever in search was the backlink: get other sites to link to you and you rank. As more answers are generated by AI, a reasonable question is whether the same signals still matter. Ahrefs set out to measure it directly. Its team looked at 75,000 brands and checked how strongly a range of factors correlated with how often each brand was mentioned in Google’s AI Overviews.¹

A correlation runs from 0 (no relationship) to 1 (the two move in lockstep). Here is how every factor ranked. Use the toggle to group them by type of signal.

Grouped by type, the pattern is hard to miss. The blue bars, signals about a brand being talked about, sit at the top. The classic link metrics cluster in the middle and lower half. Raw number of backlinks, the metric many SEO programs still report on first, correlates at 0.218: about a third as strongly as web mentions.

What counts as a “branded web mention”?

A branded web mention is any time your brand name appears on another website, whether or not it links to you. A product roundup that names you, a forum reply recommending you, a news story quoting your CEO: each one counts. A branded anchor, the second-strongest factor at 0.527, is a link whose clickable text is your brand name.

This difference makes intuitive sense for AI. A language model learns about a brand from the text around it. A page that discusses your brand in context teaches a model far more than an unlabeled link in a footer.

What does 0.66 vs. 0.22 actually look like?

Correlation numbers can feel abstract. The simulation below draws 140 imaginary brands at each correlation level. On the left, brands with more web mentions reliably tend to have more AI visibility. On the right, the cloud is so loose that knowing a brand’s backlink count tells you little about where it will land.

AI visibility is a winner-takes-most game

The study also split brands into four groups by how many web mentions they had. The gap between the groups is dramatic.

This shape matters for planning. Visibility does not grow in a straight line with effort. Brands in the bottom half are effectively invisible, and the payoff concentrates among brands that are discussed widely enough to cross into the top group.

Correlation is not causation

Two honest caveats. First, these are correlations: brands with lots of mentions also tend to be bigger, older, and better known, and some of that may drive both. Second, the study covers Google’s AI Overviews specifically. Other engines weigh sources differently, as our analysis of each engine’s citation fingerprint shows. Still, the direction is consistent with how language models work: they learn about brands from text that talks about them.

How to find your own mention gap

You do not need special tools to get a first read on where you stand. A simple audit takes an afternoon.

1. List the questions buyers ask

Write down the ten to twenty questions a buyer in your category would ask an AI assistant, such as “best tools for X” or “X vs. Y.” These define the conversations you want to be part of.

2. Collect the pages that answer them

Look at the sources AI engines cite for those questions, plus the top search results. Note every page: roundups, review sites, forum threads, comparison articles, videos.

3. Count who gets named

For each page, record which brands it mentions. Tally the totals for you and your top competitors. The difference between your count and the leader’s is your mention gap, and the pages where competitors appear without you are your target list.

Repeat the count every quarter. Because the strongest signal in the study is how widely a brand is discussed, a shrinking gap is one of the clearest leading indicators you can track.

Frequently asked questions

Sources

1. Louise Linehan and Xibeijia Guan, “An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied),” Ahrefs, May 26, 2025. ahrefs.com