Advertisers never buy search queries. They buy keywords, and Google or Microsoft decides which real searches are close enough to trigger an ad. The search terms report is the only routine look at that decision – and it is getting harder to read.
From what users typed to what Google thinks they meant
A search term is the string someone enters. The keyword is what you add to an ad group. That gap is where wasted spend and new opportunities hide.
The bigger shift is that Google now reports inferred intent for searches on Lens, AI Mode, AI Overviews and autocomplete. Google’s documentation says the term shown “represents the best approximation of the user’s intent.” For Chris Chambers, head of paid search at Understory, negative keyword work becomes
“playing whack-a-mole with shadows.”
How much spend goes dark
Google restricted the report in September 2020 to terms searched by a significant number of users. Seer Interactive measured the drop: visible search spend fell from 98.7% to 71.0% in one day, and click visibility from 98.3% to 77.9%. Performance Max initially arrived without cost-level search term data. In August 2026, Doctor Ads analysed 88 accounts and found the median account put 44.0% of clicks in the “Other search terms” row of the category insights report, versus 29.9% pooled across all clicks. That is category-level data, not the classic search terms report, so it cannot be read as cost.
Mark Meyerson of One Egg Digital earlier found about 40% of clicks and spend hidden across his agency’s accounts, and up to 85% on some keywords. Brad Geddes of Adalytics has documented identical queries showing under both exact match and AI Max. Google Ads Product Liaison Ginny Marvin responded that Google is “increasingly determining relevance by inferred intent.”
What to do with the report you still have
- Sort by cost and add negatives where queries spend without converting.
- Promote converting queries to their own keywords for bid control.
- Turn on the keyword column and treat match type labels as Google’s classification, not an independent audit.
- Audit the hidden share monthly: WordStream’s 15,666-account study associated negative-keyword use with 13% conversion versus 4.6%.
AI Max keeps redefining the boundary. It became a distinct match type in search terms reports in July 2025, aggregate rows followed in September 2025, and from June 2026 the landing page view is the default. Smarter Ecommerce found about 29% of exact match impressions in AI Max campaigns were expanded matches by July 2026, up from roughly zero in early 2025.
Microsoft Advertising reports the same data as search query performance, splitting BidMatchType from DeliveredMatchType. That distinction is useful: the match type you bid on may not be the match type that actually served.
The report is still useful. But it is now a filtered, thresholded and sometimes inferred account of what you paid for. The team that audits that gap wins back budget the team that trusts the dashboard does not.
Source: PPC Land
