Semrush study finds most ChatGPT topics lack a dominant brand
Semrush has published a guide to topical authority that draws on a study of 50,000 brands in ChatGPT, finding that 53.7% of the categories analysed had no brand appearing in at least three of five related prompts. Related prompts often returned entirely different names, so tracking a single prompt can suggest a brand is winning when the wider topic tells another story.
The guide defines topical authority as how strongly search engines and AI systems associate a brand with a subject compared with its competitors, built through a brand's own site and through mentions in reviews, publications and communities. Publishing more pages does not achieve it: brands that appeared in only one of a category's five representative prompts were linked to a lower share of brand mentions, and the negative association disappeared once a brand appeared in at least three.
Semrush's analysis also found that the sources AI systems draw on shift by subject. In research by Kevin Indig and Amanda Johnson, competitor websites accounted for 33.5% of AI citations for invoicing questions against 7% for questions about starting a business. Appearing across more of a topic's prompts was associated with stronger citation share in finance and real estate, but not reliably in legal and healthcare.
Branded search volume was the only domain-level metric that consistently aligned with topic leadership, and even that edge was modest. Google's guidance for AI Overviews and AI Mode states that established SEO practices still apply, with no separate technical requirements for appearing in its AI features.
Reported by Semrush Blog ↗
Topics: ChatGPT, AI Overviews, Measurement
In this directory: Semrush AI Visibility Toolkit