Google Search Console logs conversational follow-up questions from AI Mode, Google’s conversational search experience, as new, standalone queries. They land in the general Search Performance report, not the dedicated Generative AI report. John Mueller confirmed the mechanism on LinkedIn after marketer Anastasia Kourou posted screenshots of query strings sitting inside her account’s data: “yes,” “yes go on,” “yes, pricing.” Search Engine Roundtable’s Barry Schwartz reported the exchange on August 6.
The Generative AI performance report was built specifically to isolate AI Overviews and AI Mode activity. It carries no query-level data at all, which is why Kourou found the fragments strange enough to post about. Mueller replied with a link rather than a plain explanation. He wrote that Search Console folds AI Mode and AI Overviews data into the general performance report, then pointed to Search Console documentation for more detail. That documentation states the mechanism directly: a follow-up question inside AI Mode counts as a new user query. Its impression, position, and click totals are credited to that new turn, not to the question that started the exchange.
That framing changes how a query report should be read. A single word like “yes” carries no keyword intent on its own. Nothing about it looks worth optimizing toward. Anyone auditing the Search Performance report without this context would reasonably treat those rows as junk, sampling errors, or bot noise, and filter them out before running any real analysis. Read against Mueller’s explanation, the same rows become evidence of a multi-turn AI Mode conversation that the report cannot reconstruct in full.
The Generative AI report was built to separate AI Overviews and AI Mode results from classic organic search. It does that job for impressions and clicks. It strips out the query string entirely, the one field that would let a team confirm a session actually came from AI Mode rather than guess. The general Search Performance report keeps the query string, but it mixes AI Mode turns into the same list as every other organic search, with no filter or label to split them apart. Query-level visibility into AI Mode exists in exactly one place, and that place was not designed for the job.
The SEO community has been working around this gap informally, scanning for unusually long, conversational query strings as an informal signal of AI-driven traffic. Ross Tavendale raised the identical problem in a post on X on August 4. He described Search Console as recording AI Mode responses and asked how to reverse-engineer the pattern. Mueller’s reply gives that guesswork a documented basis, without adding any new reporting capability to act on it.
Search teams reviewing this week’s query data should stop discarding one-word and command-style entries such as “yes,” “go on,” or “sure” as noise. Check first whether they cluster around known AI Mode-heavy topics before filtering them out. Because the general report is the only source with query-level detail, any count built from it should be treated as an undercount of AI Mode volume, not a clean measure of it, until Google adds a way to separate the two directly.
Search Engine Roundtable, reported by Barry Schwartz on August 6, 2026.
