Google AI Max for Search What DTC Brands Should Expect Before Switching It On

Every DTC operator running Google Ads has now had the same conversation with the same rep. Switch on AI Max, get 14% more conversions at a similar CPA, do nothing else. It is a compelling pitch because it asks for one click and promises free growth.

The 14% figure is real. It also explicitly excludes retail. Once you look at what happened in actual ecommerce accounts, the picture changes shape, and the decision stops being about whether AI Max works and starts being about whether your margin can absorb the way it works.

What AI Max actually is

AI Max is not a new campaign type. It is a bundle of opt-in features that sit inside a standard Search campaign you already run. The headline component is keywordless matching, which lets Google serve your ads against searches that appear nowhere in your keyword list. Alongside it you get automatically created assets and dynamic final URL selection, which sends traffic to whichever page on your site Google decides fits the query best.

That distinction matters more than the marketing does. Performance Max takes your budget into a black box across seven surfaces and gives you almost nothing to diagnose with. AI Max leaves you inside Search, with your keywords, your negatives, your search terms report and your bid strategy intact. You are widening the funnel, not surrendering the account.

For a brand that has spent two years building a clean Search structure, that is the difference between a test and a gamble.

The gap between the pitch and the data

Google's own benchmark is 14% more conversions or conversion value at a similar cost per action, rising to 27% for campaigns still leaning heavily on exact and phrase match. Read the footnote and retail is carved out of that number. When a platform excludes your category from its own headline stat, that is the most useful thing it has told you.

An independent study of more than 250 Search accounts running AI Max found a median revenue uplift of 13%, almost exactly matching Google's non-retail figure, alongside a median CPA increase of 16%. The ROAS outcome range ran from +42% at the top to -35% at the bottom. That is not a small variance. That is two entirely different businesses depending on which side of the median you land.

The same analysis found that up to 63% of the time, the queries AI Max surfaced through keywordless expansion were recycling coverage the account already had. Searches that existing keywords were already capturing were simply being matched through a different mechanism and reported as incremental discovery. You pay more for the same customer and the interface congratulates you.

The brand cannibalisation problem

This is the criticism practitioners raise most, and it is the one that does the quiet damage. Keywordless matching expands into queries you would rather it left alone, including competitor brand names and, more often, your own. If you run a separate branded campaign, AI Max starts eating it.

What that does to your reporting is worse than what it does to your spend. Branded searches convert at a far higher rate than cold ones, so the prospecting campaign that swallowed them looks like it just found a new gear. Meanwhile the branded campaign loses impressions and looks like it is fading. You then make a budget decision based on both of those readings, and both of them are wrong in the same direction.

Google has since shipped native branded search controls inside AI Max, with three settings that let you decide whether ads serve on queries containing brand names. If you run a separate branded campaign, set it to unbranded only before you switch anything on. Doing it afterwards means your first four weeks of data are unreadable.

What this looks like in practice

Test it on one campaign, never the account. Pick a non-branded campaign with enough volume to reach significance inside a month, and leave everything else exactly as it was. Same bid strategy, same target, same budget. If you change three things at once you learn nothing about any of them.

Before you enable it, set the branded control to unbranded only, apply every negative keyword list the account already uses, and check the dynamic final URL setting. That last one quietly reroutes traffic to whichever page Google prefers, which on a DTC store often means a collection page instead of the product page you optimised, or a blog post instead of a product page at all. Pin it to the URLs you want or exclude the pages you do not.

Then give it three to four weeks. The learning period distorts anything shorter, and a two-week read on a keywordless expansion is a coin flip dressed as a decision. Pull the search terms report weekly during the test and tag every new query as either genuinely new demand or coverage you already had. If more than half falls into the second bucket, you have found your answer before the CPA data even matures.

Judge the result on contribution margin per order and new customer CPA, not on the conversion count in the Google interface. Platform-reported conversions include returning customers, which is precisely the population keywordless matching is most likely to hoover up. If new customer CPA rises by more than your margin can carry after four weeks, turn it off and keep the negatives you learned.

Who should switch it on, and who should not

The accounts that gain most are the ones still running tight exact and phrase match structures with a narrow keyword list, in categories where customers describe the problem rather than the product. A supplement brand whose keyword list covers product names but not the twenty ways people describe the symptom has genuine uncovered demand for the system to find. Google's own 27% figure points at exactly this group.

The accounts that lose are the mature ones. If you have already built out broad match with a strong negative list, already run Performance Max alongside Search, and already cover the long tail, keywordless expansion has very little left to discover. It reaches for the nearest available inventory, which is your brand and your competitors, and your CPA climbs to buy traffic you were getting for less.

The third group is the one nobody warns: brands with thin contribution margin. If you are running under 25% after all variable costs, a 16% CPA increase is not a trade-off to evaluate. It is a loss. Fix the margin first, then come back to the test.

Inside the system

How we build this for brands

A test like this only produces a real answer if the measurement underneath it is honest. For the brands we work with, we build profit and cash-flow dashboards from live Shopify and ad data so that every paid media decision is read against contribution margin per order rather than platform-reported conversions. A reporting agent runs against that data weekly, surfacing CPA drift, margin leakage and campaign cannibalisation before a month of spend has gone through at the wrong number.

Alongside it sits the VOC engine, which mines customer reviews and support messages for the language people actually use to describe the problem your product solves. That is the same language keywordless matching is trying to guess at, except you own it, you can put it into your keyword list, your ad copy and your landing pages, and you can see exactly what it costs you. Part of this runs live for portfolio brands today; the full system is what we deploy when we take a brand on.

Frequently asked questions