geoagencies

ChatGPT Searches the Web in Only 14% of Prompts, Ahrefs Study Finds

Photo by BoliviaInteligente, Unsplash · source ↗

Ahrefs analysed 66.6 million free, logged-out GPT-5.6 Luna responses and the 15 million fan-out queries they generated across August and September. Only 14.4% of responses triggered a web search at all, and when one did, ChatGPT ran roughly 1.6 fan-out queries on average.

Of those fan-out queries, 6.8% used a site: operator to search a single domain, spread across about 116,000 sites. Among the roughly 160 most-searched domains, 75.9% were brand sites such as apple.com or ikea.com, while retailers including Walmart, Costco and Amazon accounted for 11.9% and Reddit and Facebook just 1.9% between them.

The language inside fan-out queries shifted toward commercial intent. The word "official" appeared in 13.1% of English fan-outs in September, up from 10.1% in August, and terms like "price" and "review" rose by as much as 47% month over month. In separate citation data, ChatGPT's Reddit citations fell 88% over six months.

The practical implication for brands is that most prompts are answered from what ChatGPT already knows, so visibility depends on prior training data as much as on retrieval. Where ChatGPT does search, it favours official sources for factual questions about pricing, specifications and company details, which means those facts need to sit in plain text on a crawlable brand page.

Reported by Ahrefs Blog ↗

Topics: ChatGPT, Citations, Measurement

In this directory: Ahrefs Brand Radar

Related