ChatGPT prompts show thin purchase intent, study finds
A study of 118,163 ChatGPT prompts drawn from 41,961 anonymised conversations has put numbers on how Americans actually use AI assistants, and the picture is less commercial than marketers might assume. The data came from Verso, which makes opt-in, anonymised connections to OpenAI, and covered 12,321 panel users between 1 January and 29 October 2026. Users averaging more than 33 prompts a month were excluded because their volume skewed the sample.
Only 6.9% of prompts showed any sign of researching a brand to buy from. By comparison, more than 15% of 332 million Google queries analysed two years ago matched the same definition, which suggests people still bring purchase-journey questions to search far more often than to ChatGPT. Responses contained a brand, product or company recommendation slightly more often than users asked for one, at 12.1%.
The panel skews consumer rather than professional. Compared with OpenAI's own published usage categories, Verso's users did more practical guidance and image generation and less coding and technical help, a gap the report attributes to API traffic being absent from the panel. Tech workers in the sample were markedly more likely to ask coding questions and less likely to use the tool for homework or emotional conversations.
For brands chasing citations in AI answers, the finding cuts both ways. Brand mentions in responses are common enough to matter, but the volume of commercial-intent conversations is small next to search, so AI visibility work sits alongside search rather than replacing it.
Reported by SparkToro ↗
Topics: ChatGPT, Measurement, AI advertising