Keywords were once the closest thing paid search had to a contract. You picked a term, assigned a match type, and understood which searches could trigger your ad. That foundation is being rewired around AI-driven matching, and Google’s September 2026 migration schedule makes the change hard to ignore.
What changed
On 3 August 2026, Google stopped allowing new campaign-level broad match and automatically created assets configurations. On 1 September, it began converting those campaigns to AI Max. Dynamic Search Ads were initially scheduled for the same treatment, but Google pushed that automigration to February 2027 after pushback from practitioners.
Exact match is no longer fully literal either. Since 30 August 2025, Google’s enhanced AI keyword matching has interpreted exact match on semantic meaning rather than exact words. Testing from Smarter Ecommerce found exact match losing roughly one impression in four to AI Max. In separate independent analysis across about 30,000 search terms, fewer than half showed any keyword match at all.
Reporting adds another layer of confusion. Since May 2026, the search terms report has shown inferred intent rather than the query someone actually typed. With AI Max active, traffic can be attributed back to the narrower match types in the account, so a single exact-match row may blend several delivery paths.
Why advertisers should care
The keyword is still where budgets are planned, forecast and audited. But precision has slipped. WordStream benchmarks across 13,474 campaigns put the average Google Ads cost per click at $5.42 for 2026, up from $2.32 in 2016. Paying more per click is harder to justify when you cannot see exactly which query triggered it.
Negative keywords behave asymmetrically. They do not expand to close variants: excluding flowers blocks red flowers but can still serve on red flower. Worse, because AI Max works on inferred meaning while negative lists match text strings, an exclusion can be routed around.
Many accounts were built by isolating every match type into separate campaigns and ad groups. Google has actively discouraged that practice for years because the systems now read synonyms and intent. Maintaining exhaustive segmentation is often more costly than the control it appears to offer.
A practical reset for growth teams
- Consolidate redundant match-type silos. If broad, phrase and exact variations serve the same intent, merge them and let AI Max handle query-level routing.
- Re-audit negative lists. Text-string exclusions are not enough; use brand controls inside AI Max where available.
- Treat exact-match reports as multi-source, not literal. Segment by campaign and landing page rather than trusting a single row as user input.
- Stress-test forecasts. Keyword Planner now offers city, region and device breakdowns, but AI Overviews are already reducing click-through on some terms.
What this means for your strategy
Keywords are not disappearing; they are being folded into automation. Search demand still carries commercial intent, and a keyword remains the basic lever for buying that demand. The change is that the matching engine now decides more of the reach than your punctuation does.
Teams that still treat exact match as literal control will keep fighting the interface. The better move is to manage the inputs that compound: creative relevance, landing page experience, query exclusions, and a structure that gives AI Max useful signals instead of microscopic pegs.
Source: PPC Land
