The Report Looks Like a Dumpster Fire. The Revenue Says Otherwise.
Here is a scene we see on e-commerce accounts constantly. You (or your Google rep) turn on AI Max. A week later you open the search terms report and it is full of junk. Searches that have nothing obvious to do with your product. A furniture brand that only sells American-made pieces showing up for "buy England furniture online." A store matched to the single word "home." Your first instinct is to start negating everything in sight.
Then you look at the revenue. And sales are up.
Now you are stuck between two honest reactions, and both are reasonable. The first is "great, it is working, leave it alone." The second is "wait, is that lift actually from AI Max, or did something else happen this month and I am about to give the robot credit for it?" That second question is the smart one, and almost nobody asks it. This article is about answering both: why the junk terms can be a good sign on a store specifically, and how to tell whether AI Max earned the lift or just happened to be standing there when it arrived.
An Online Store Is Not a Lead-Gen Account
Most of the AI Max advice online is written for lead-gen businesses, and for good reason: that is where AI Max is riskiest. When a plumber's account matches a junk search, that click costs money and drives toward a form fill that may never come. Waste is waste.
An online store plays a different game, and the difference is the whole reason your sales went up. Smart Bidding on an e-commerce account is not optimizing toward a vague "maybe this becomes a lead." It is optimizing toward actual purchases and actual revenue, fed by a live signal of what really sold. So when AI Max reaches past your keywords, it is not guessing in the dark. It is chasing the searches that its data says end in a checkout, then bidding to your real ROAS target.
That flips the meaning of a weird search term. On a lead-gen account, an odd match is usually money leaking out. On a store, an odd match that still results in a sale is discovery: a pocket of demand you would never have found by typing keywords into a box. The messy term is the surface. The purchase underneath it is the truth, and the purchase is what Smart Bidding is actually steering by.
This is also why Google's own numbers skew favorable for retail. Google says advertisers who turn AI Max on typically see about 14% more conversions or conversion value at a similar CPA or ROAS, and up to 27% for accounts still leaning on exact and phrase keywords. In one of Google's own e-commerce examples, L'Oreal reported roughly double the conversion rate at a 31% lower cost per conversion. Take the vendor's figures with the usual grain of salt, but the direction matches what we see: on a well-tracked store, AI Max tends to find real buyers.
Why the "Junk" Converts (You Are Seeing One Signal Out of Fifty)
Here is the part that makes account managers uncomfortable. When you look at a search term, you see a search term. That is it. When Google decides whether to show your ad for that search, it is weighing dozens of other signals you never see: what that person has bought before, what is sitting in their cart, what they looked at an hour ago, their device, their location, the time of day, where they are in a buying cycle. The search term is the tip of the iceberg. The intent that made the bid worth it is underwater.
The agency RoiMonks put this well after a five-month, 20,000-dollar AI Max test: do not rush to exclude generic terms that are still even loosely relevant, because AI Max is reading conversion-probability signals you cannot see, while all you have is the words. Their rule, and ours, is simple: only negate what is 100% irrelevant. "Buy England furniture online" for an American-made brand? Cut it, it is nonsense. A broad, generic term that keeps quietly producing sales? Leave it alone, even if you cannot explain why it works.
The second reason the junk converts is that shopping itself has changed. Google now says the large majority of searches are longer, more conversational, and more specific than the tidy two-word phrases we built keyword lists around, and a growing slice come from image and voice search rather than typing. People do not shop in keywords anymore. They shop in messy, human sentences. A lot of what looks like garbage in your report is just the actual, unpolished way real customers now search for things to buy.
"Sales Went Up" Is Not the Same as "AI Max Did It"
Now the hard part, and the reason your instinct to question the lift was correct. A rise in sales after you flip a switch is not proof the switch caused it. Before you credit AI Max, you have to rule out three ways it can look like a hero without being one.
First, brand terms. If AI Max is allowed to match searches for your own brand name, it will scoop up people who were going to buy from you anyway, then take credit for those sales. Suddenly AI Max looks brilliant. Fix: in any non-brand campaign, add your brand as a brand exclusion so AI Max has to prove itself on genuinely new demand, not on customers who already knew your name.
Second, cannibalization. AI Max can lift its own numbers by quietly stealing conversions from your other campaigns, your Shopping, your Performance Max, your existing search. The account looks flat or up a little, but AI Max's slice looks amazing because it ate everyone else's lunch. Fix: watch total account performance and each other campaign's numbers during the test. In the RoiMonks test, the tell that AI Max was genuinely additive was that the other campaigns held steady or improved slightly while AI Max grew. It added instead of reshuffling.
Third, the calendar. Maybe you turned AI Max on the same week a promo landed, a season turned, or a competitor went dark. The only clean way to separate AI Max from the calendar is a holdout: run a 50/50 experiment where half your traffic gets AI Max and half does not, at the same time. Same weather for both, so the difference between them is actually AI Max and not the month. If an experiment is too much, at least insist that total account conversions rose, not just the AI Max line, before you declare victory.
None of this means AI Max did not earn your lift. It very well might have. It means you should be able to say why you believe it did, instead of pointing at a bigger number and hoping.
What AI Max Is Actually Doing Under the Hood
Strip away the branding and AI Max is three levers bolted onto a normal Search campaign. Search term matching opens your keywords up to broad match plus keywordless technology, so Google can reach searches your keyword list never named. Text customization lets Google write and rework headlines and descriptions by reading your website. Final URL expansion lets Google pick which page on your site to send each click to, rather than everyone landing on the one URL you set.
Google's own product team describes AI Max as essentially "Performance Max for Search," with one crucial difference for people who like control: it stays grounded in your keywords and ad groups instead of going fully keywordless like PMax. You keep more of the steering wheel.
Here is the counterintuitive part, and it matters for judging your own results. If your store already runs broad match and dynamic search ads, AI Max may not actually expand your reach much at all. The RoiMonks test found AI Max was bidding on roughly the same searches they were already targeting. So where did the extra conversions come from? Higher ad relevance. By auto-writing a tighter ad and picking a better-matched landing page for each search, AI Max raised Quality Scores and Ad Rank, which won more of the auctions the account was already competing in. Not new territory. The same territory, won more often. That is a very different, and more believable, story than "the AI found magical new customers," and it is worth knowing which one is driving your numbers.
Your Product Pages Just Became an Ad Lever
Because text customization and final URL expansion both read your website, AI Max quietly promotes your product and category pages into part of your ad account. Google writes your headlines from your page copy and chooses your landing pages from your catalog. A thin, messy, or outdated product page now directly drags down the ads Google builds from it.
That cuts two ways for a store. The upside: tighten your product page titles, descriptions, and structure and you improve the raw material AI Max has to work with, which lifts the ads themselves. The watch-out is Final URL expansion. Left unchecked, it can send a shopper searching for one product to a page for a different one, or to a weak page that converts badly. Check the AI Max landing-pages report, and if Google keeps steering clicks to pages you do not want, use URL exclusions to fence those off, or turn Final URL expansion off for that campaign until your pages are ready. On a store, the page is no longer just where the click lands. It is a targeting and copy input.
How to Test It Without Lighting Money on Fire
If you have a real store, clean conversion tracking, and a budget you can afford to bruise, AI Max is worth a controlled test. If you are missing any of those three, fix that first, especially the tracking, because everything Google optimizes toward is only as good as the purchase data you feed it.
Keep the test contained. Turn AI Max on for one campaign, not the whole account, and use a slice of budget you would not miss if the first few weeks are ugly, because they will be. Every account we have tested, and every one the specialists describe, goes through a rough patch first: high CPCs, weird matches, conversion value below cost while Google learns. The RoiMonks account took about six weeks before it turned profitable, then beat the rest of their non-brand search in months three through five. If you panic and kill it at day five, you have paid the tuition and skipped the lesson. Give it around two months, and judge it on the trend, not on Tuesday.
While it runs, work the AI Max search-term and landing-page report weekly. Negate only the genuinely irrelevant. And do the move most people miss: when AI Max surfaces a search theme that is quietly converting, build it into your own campaigns as a real keyword or ad group. Used this way, AI Max is not just a campaign setting, it is a discovery engine that shows you demand you can then capture deliberately. If it flat out does not work for you now, retest in six to twelve months. These tools improve fast, and there is a deadline coming anyway: Google is upgrading dynamic search ads into AI Max through 2026, so if your store leans on DSA, this is a shift you will be making regardless.
The Honest Verdict for a Store
For a small lead-gen business on a tight budget, our advice on AI Max has been "not yet." An online store is where that answer changes. The economics are different, the junk terms are often real discovery instead of real waste, and Google's own retail numbers lean positive. If your tracking is clean and your product pages are in good shape, AI Max deserves a genuine, contained test.
Just hold both truths at once. Yes, those ugly search terms can be exactly how modern customers shop their way to your checkout. And no, a good month is not proof until you have ruled out brand credit, cannibalization, and the calendar. Give the tool a fair test, then make it show its work.
If you would rather have someone run that test properly, watch the search terms, protect against the false positives, and tell you the honest answer, that is what we do all day at BrandRocket. Either way, now you know what your store's search terms report is really telling you.
