Google's Ginny Marvin has set out post-GML priorities, and the keyword is being demoted. AI Max for Search uses Gemini to match ads on inferred intent, with first-party data and measurement at the centre. It rewards advertisers who know their numbers and exposes the ones flying blind.
Google is steering you away from controlling the keyword and toward trusting the model. That is a fine trade if you have the data to keep it honest, and a leap of faith if you do not.
Google's ads product liaison Ginny Marvin has spelled out what advertisers should focus on after Google Marketing Live, and the direction is unmistakable. AI Max for Search, the new layer that applies Performance Max-style AI to standard search campaigns, is now rolling out globally with full availability expected by early in the third quarter.
What it does is the headline. AI Max uses Google's Gemini model to match your ads to inferred user intent rather than the exact keywords you typed in. The machine reads what the searcher means, not just what they wrote. Marvin framed the wider push around measurement and first-party data, and the strategic message was clear. Google wants advertisers optimising toward predicted business outcomes, not raw conversion volume.
That is a quiet but real shift. The keyword, the thing search advertising was built on for two decades, is being demoted to a signal among many.
Why it matters
Handing intent matching to Gemini means handing Google more of the decision about who sees your ad. The payoff is reach and efficiency. The cost is control, and the only thing that buys back control is your own data. Marvin putting first-party data and measurement at the centre is not a throwaway line. Without it you are trusting the model with no way to check its homework.
Full global rollout of Google AI Max for Search is expected by early in the third quarter of 2026
This rewards the businesses that already know their numbers and own their data, and exposes the ones flying blind. If you cannot measure what a conversion is worth, optimising toward predicted outcomes is optimising toward a target you cannot see.
What to do about it
Search advertising is becoming a conversation between your data and Google's model. The advertisers who own clean data and know their numbers get the upside. The ones who do not are handing the keys to a machine they cannot question.