Unlinked brand mentions

Google's John Mueller stated in 2022 that an unlinked mention passes no signal the way a link does, since there is nothing for the system to attribute credit to^[1]. That reasoning describes the link graph. In Ahrefs' data the ordering is reversed for AI answers: branded web mentions correlate with AI Overview visibility more strongly than any backlink metric tested^[3].

In the same 75,000-brand analysis, branded anchor text scored 0.527 and branded search volume 0.392, both ahead of backlink count and Domain Rating^[4]. A later roundup put brands in the top quartile for web mentions at 10 times the AI visibility of the rest^[5]. These are correlations on observational data, so a large brand being mentioned often and cited often is the obvious confounder.

Mention volume matters because recognition is the input a recommender works from. A controlled experiment on skincare products across GPT-4o-mini, Claude Sonnet and Gemini 3 Flash found well-known brands recommended 100% of the time when every competing product carried the same specifications, and that dominance disappearing as soon as a rival held a rating advantage of less than 0.1 stars^[6]. Recognition sets the default answer and a small verifiable difference overturns it, which is the mechanism the correlations above are picking up.

A larger audit measures the same ladder from the other end: about 37,000 production runs, 215 commercially framed prompts across 19 sectors, and a 533-brand catalogue sorted into five prominence tiers from category leaders down to regional players^[7]. The tier-by-tier results sit under AI visibility measurement.

The May 2025 study tested eleven brand and site factors and YouTube mentions were not among them^[8]. A later Ahrefs analysis of ChatGPT visibility found YouTube mentions and YouTube mention impressions the strongest correlate of any factor tested, ahead of branded web mentions^[9], which suggests the surface a mention appears on matters as much as the count. See ChatGPT Search.

Being cited and being named are separate outcomes. Across 3,981 domain appearances spanning 115 prompts and 14 countries in ChatGPT, Gemini, AI Overviews and AI Mode, 74.9% were citations but only 38.3% named the brand in the answer text^[10]. That leaves 61.7% as ghost citations: the domain supplied the source link and the reader never saw who wrote it^[11]. A brand measuring only link-outs is therefore counting a different thing from a brand measuring recall.