Should You Use Google AI Max for Search?

Your Google rep has emailed twice this month about AI Max. The deck says "up to double-digit conversion lifts," the call ends with "we can enable it on your account today," and you are the one who has to explain the spend report next quarter. You have been here before with Performance Max: more conversions on paper, a search terms list full of queries you never wanted to pay for. So the question is not whether AI Max is clever. It is whether flipping it on costs you the control you spent two years building.

Short answer: Google AI Max is worth testing, not flipping on blind. It bundles three separate features (broader search-term matching, generative ad text, and dynamic landing pages) onto a Search campaign. Trial it on one campaign with your negatives, brand exclusions and a tight conversion goal already in place, then judge it on cost per qualified sale.

The takeaways

• AI Max is three features in one switch: search-term matching (broad match plus keywordless), text customization (generative RSAs), and final URL expansion. You can turn each off.
• The independent read is a coin toss: a 250-campaign retail study by Smarter Ecommerce found a median +13% revenue but a median +16% CPA, with ROAS ranging from +42% to -35%.
• Final URL expansion is on by default. If only specific landing pages are approved, uncheck it before you launch.

What does Google AI Max actually turn on?

Google AI Max is not a single setting. It is a bundle of three features you enable together on an existing Search campaign, and per Google Ads Help each one does a different job. Search-term matching uses broad match and "keywordless" technology to reach queries beyond your keyword list. Text customization pulls from your ads, landing pages and generative AI to rewrite headlines and descriptions per search. Final URL expansion sends a click to whichever page on your site the system judges most relevant, instead of the URL you set. The reason the "is AI Max good" debate goes in circles is that people argue about the bundle when the risk lives in the individual parts. You can enable search-term matching and leave final URL expansion off. Treat it as three separate decisions.

Where does AI Max help, and where does it widen spend?

AI Max helps most where your keyword list is the bottleneck: it finds net-new, relevant queries a human would never think to add. That is real, and for accounts capped by a thin keyword set it can open genuine volume. The cost is the same mechanism running the other way. Broad matching into queries beyond your list is exactly how budget leaks into junk, the Performance Max problem in a new coat. The independent numbers say be careful: Smarter Ecommerce's study of 250-plus retail campaigns found a median revenue lift of 13% but a median CPA increase of 16%, and ROAS swung from +42% to -35% depending on the account. That spread is the headline. AI Max is not a free win or a scam. It is a high-variance bet whose downside is real, so you test it, you do not trust it.

How do you trial AI Max without handing over the account?

Arm the guardrails before you flip the switch. The first bad week is too late. Standard negative keywords still apply in AI Max and are enforced before the expansion logic runs, so load your existing negative lists and add competitor brand terms first. If AI Max sits on a generic campaign, add a brand exclusion for your own brand so it hunts net-new queries instead of cannibalizing your branded exact-match campaign. Uncheck final URL expansion unless every page on your site is a fair landing page for a paid click. Point the campaign at a real conversion goal such as a purchase or a qualified lead instead of a raw form-fill, so the algorithm optimizes toward money and not volume. Then read the search terms report every few days for the first two weeks. The report is your evidence. If the new queries are junk, the negatives go in and you have your answer.

Should you enable it because your rep asked?

No. A rep's quota is not your profit, and "we can turn it on today" is a sales close. You do not have to accept it. That does not make AI Max wrong for you, it makes the decision yours. Run it as a deliberate test on one campaign, with a clear conversion goal and a fixed window, and compare cost per qualified sale against the campaign it replaced. This is the same discipline good automation runs on everywhere: a rule that fires without a human keeping veto over it is how accounts quietly break. It is the logic behind how Adscalr's automation layer works too, where the rules only pause or kill, never scale on their own, and the default is a recommendation a human approves. Let the test tell you whether to keep it on. If you want the wider version of this fight, the Google Ads recommendations tab plays the same game with the auto-apply toggle.