ChatGPT and Google AI Mode name the same brands twice as often as they cite the same sources

The 67% and the 30% do not compare a brand name against its own website. They compare two systems to each other: of the brands surfaced across ChatGPT and Google AI Mode, 67% appear in both, and of the cited sources, about 30% do ^[3]. Two retrieval pipelines can converge on a brand list while drawing on largely separate documents. Divide one figure by the other and the stacks agree on names roughly 2.2 times as often as they agree on evidence ^[1].

Per-response citation counts decide what one bad page costs. Semrush puts ChatGPT at about 15 sources per answer and Gemini at around three ^[4]. On a three-source answer, a single page is about a third of the cited evidence; on a fifteen-source answer it is about 6.7% ^[3]. The same outdated spec table therefore weighs roughly five times more on the platform that cites less ^[2]. Semrush says neither figure establishes that one platform is inherently more trustworthy ^[8].

Both overlap figures are aggregates over 126 million prompts ^[1]. For 30% to describe a particular category, that category's prompt mix would have to resemble the index's mix. The write-up says a buyer asking for product comparisons, local options, implementation guidance or expert recommendations may get answers built from very different types of material ^[12], and it argues that a detailed specialist publication can be more useful for a technical question than a broad directory that appears more often elsewhere ^[15].

Semrush reports that mentions and citations often do not match ^[2]. A company can be named in an answer while its own website is absent from the cited sources, and a third-party review, publisher or retailer may help define how that company is described ^[10]. The write-up does not say how often ^[14].

Domain-level strength still matters here. The same write-up says authority-style thinking remains relevant to many search strategies, and adds that models assemble responses from source sets that change by platform and prompt ^[9]. It also notes that an AI answer can influence research and purchase decisions without sending the user to the company's site, and that absent, outdated or inconsistent independent sources can leave a strong owned-content program short of controlling the description an AI system assembles ^[13].