Google Ads search terms from AI Mode and AI Overviews may show inferred intent, not what users typed

Google has quietly updated a Google Ads help page to clarify that the search terms shown in reports for AI-powered search experiences may not match the user's exact query. In some cases, the term displayed represents Google's interpretation of what the searcher intended. The clarification applies to AI Mode, AI Overviews, Google Lens and autocomplete.

According to Search Engine Journal, in a report published on November 9, the new language appears in the documentation on ad group prioritization. That is the page explaining how Google decides which ad group enters the auction when multiple keywords or targeting methods could match the same search. The change was first spotted by Anthony Higman, who shared the find on LinkedIn.

In practice, advertisers may find terms in their reports that a user never typed, appearing as a normalized or interpreted version of the interaction. The search terms report never offered full visibility, but it was common to assume that a term listed there was the actual query.

Why Google would change the report

The report suggests the update reflects how hard it is to report on AI experiences, especially after the recent announcement of more ads in these environments. Users refine a search across several prompts, search by image, or accept autocomplete suggestions before finishing a query. Sometimes there is no single, clean query to show.

There is also likely a privacy component, according to the analysis. As search becomes more conversational, people provide more context, and Google may not want to expose every raw prompt, image search or refinement in advertiser reports.

What remains unclear

Google has not publicly explained how much interpretation takes place, whether advertisers can tell modeled terms apart from literal queries, how negative keywords behave against interpreted intent, or how close the terms stay to the user's original wording.

The most exposed are regulated industries, which review search terms for compliance and brand safety; B2B advertisers, who use queries to map customer pain points; and ecommerce, which builds negative keyword lists from them. Accounts already optimizing around intent themes, with broad match and Smart Bidding, are likely to notice less of a difference.

For brand teams, the report's takeaway is to treat the search terms report as a signal of the direction of intent, not as the customer's exact language. That calls for more caution when presenting these terms to executives and clients as proof of what consumers are looking for, and it is likely to shift optimization toward signals such as landing page alignment, first-party data, conversion quality and content relevance.