ChatGPT launched nearly four years ago, and Claude followed a few months later, yet detailed data on consumer prompts and platform response methods remains scarce. Researchers rely on traditional search query data to infer genAI interactions, with only Bing offering AI transparency through reporting the prompts, termed “Grounding Queries,” that drive “fan out” responses. Now it is known that Google’s AI Mode records initial and follow-up prompts as search queries.
An SEO Manager in Greece, identified as Anastasia Kourou, tagged Google’s John Mueller on LinkedIn asking why Search Console listed genAI-like prompts such as “yes” and “yes, pricing” as search queries. Mueller confirmed these were follow-up prompts from users on AI Mode, validating the Search Console data.
To identify potential AI prompts in Search Console, users can filter long queries using regular expressions. Starting in the “Performance” section, go to “Add filter.” Then choose “Query” > “Custom (regex).” Then type: You’ll see long, prompt-like queries: An SEO strategist at Tripadvisor, Jean-Christophe Chouinard, posted a more detailed regex that includes follow-up prompts, such as “yes, please” and “tell me more.”
Read Also: Beauty Brands Gain Retail Access Through Spa Partnerships
Search Console’s API provides more data than the front-end interface, and third-party tools can access this data without requiring developer involvement. Two free tools are commonly used: Queries, which extracts prompt-like AI Mode queries from Search Console, allowing categorization by topic and intent, blocking of likely bot prompts, and exporting to spreadsheets; and Search Analytics for Sheets, a Google Sheets extension that exports Search Console data, groups queries using the regex method, and pulls up to 25,000 queries with built-in Gemini capabilities for analysis.
Search Console does not definitively label which queries are prompts, but the wording often makes them identifiable. Some long queries show high impressions, possibly from prompt-tracking software rather than human users, though this still provides value by revealing competitor monitoring activities. Prompts tend to be detailed and unique, not fitting standard categories or metrics, and practitioners are still developing ways to interpret this data for optimization strategies.
Using ChatGPT, Gemini, or Claude to analyze and categorize prompts can help align content with user needs. Focusing on target customers and identifying their pain points allows marketers to anticipate likely prompts.
