Google's help page for AI Max for Shopping gives a simple example of what it does. A product titled "Mens Shoe SKU 123" might show to shoppers as "Men's Waterproof Hiking Boot." Google's AI reads your feed and your website, then rewrites the title to match what the shopper searched for.
For a store with weak titles, that sounds like free help. And Google builds in a safeguard: a rewritten title only runs if Google predicts it will do better than yours. If it doesn't, your title runs as you wrote it.
We still turn AI Max on last. It changes two things a store has spent the whole guide getting right: the words in its ads and the pages its clicks land on. This chapter covers what AI Max for Shopping rewrites, the claims it can add, what Final URL expansion does, what Google promises, and the order of steps we finish before switching it on. The examples come from the cookware stores in this guide. Your products will differ. The order of steps holds.
Google Turns "Mens Shoe SKU 123" Into "Men's Waterproof Hiking Boot." When Google Predicts No Lift, Your Title Runs as You Wrote It.
Google describes AI Max for Shopping as a "beta suite of features" for Shopping campaigns. Its first feature, text customization, rewrites product titles. Google says it uses "content from your Merchant Center feed" and "website content while using generative AI" to create titles matched to each search.
Google lists two guardrails. Titles are checked by quality models meant to keep them "grounded in your actual product offerings." And, in Google's words, "Customized titles will only serve if they're predicted to perform better than the original titles you uploaded. If the system does not predict a performance uplift, your original title will serve."
Two limits matter for a store today. Text customization is "currently limited to English feeds." And rewritten titles can show beyond the Shopping row; Google names AI Overviews among the newer places they can appear.
Notice what that example really shows. "Mens Shoe SKU 123" is a bad title. A store whose titles already name the product type, brand, material and size, the way Chapter 14 will describe, leaves AI Max much less to improve. In our view, the rewrite likely helps most where the feed is weakest, which is exactly where a store should fix the feed first.
Google's Own AI Max Example Adds "Made in USA" to a Sofa Title. The FTC Holds a Store to Those Three Words.
Google's help page includes a table of titles customers might see after rewriting. One of them reads "Made in USA 93" Deep Sofa."
That's three words with legal weight. The FTC's Made in USA Labeling Rule says a product can't be labeled made in the United States unless its final assembly or processing happens in the U.S., all significant processing happens in the U.S., and "all or virtually all ingredients or components of the product are made and sourced in the United States." The rule covers mail order and online promotional material, and the FTC can seek civil penalties for an unqualified Made in USA label on a product that doesn't qualify.
Google's example is a sofa, and it's an illustration of a rewritten title, not a claim about any real product. But it shows that text customization can put an origin claim in an ad. If that claim is wrong for your product, it's your ad.
Store titles already carry strong claims without any help from AI. In our pulls of cookware Shopping ads, Caraway titles stacked "Non-Stick, Non-Toxic, PFAS-Free," Misen put "2026 Best" in front of nine product titles, and a Made In title included "USA-Made." Each of those can be true and properly supported. Each is also the kind of word that needs checking before Google repeats it, or adds it somewhere new.
The campaign settings give a store three ways to fence this in. A list of blocked words keeps specific phrases out of anything Google writes, up to 25 per campaign, with capital letters ignored. A separate set of up to 40 written instructions tells Google which ideas, associations or tones to stay away from. And brand and page exclusions keep the campaign off names and URLs you choose.
We fill the term exclusions before AI Max goes live, starting with any claim the store can't prove for every product: origin, "non-toxic," "best," warranty terms, certifications. A word the store can prove can stay. A word it can't goes on the list.
Final URL Expansion Sends Shopping Clicks to Pages Google Picks. A Store Can't Use It Until Target ROAS Is Running.
AI Max for Shopping's second feature reaches past the Shopping row. Google says Final URL expansion "advances eligibility beyond basic Shopping ads to text ads," using "text ads powered by your Merchant Center feed and dynamic landing pages."
To pick those pages, Google says it "crawls your website to identify other high-value 'commercial URLs', for example, category pages, 'new arrivals' or editorial pages," and sends traffic to pages it judges "query-relevant and themed to your ad group." It comes with conditions:
- "Final URL expansion is only available with a Target ROAS bid strategy."
- It "requires text customization to be turned on," so the ad copy matches the landing page.
- URL exclusions let you keep it off pages you don't want, such as the blog, help center or policy pages.
- Campaign priority, which Chapter 6 covered for overlapping Shopping campaigns, "doesn't impact final URL expansion."
Google has also said a one-click AI Max toggle is coming to Shopping campaigns, with the option to turn Final URL expansion off at any time and keep delivery to Shopping ads.
The Target ROAS requirement puts Final URL expansion late by design. A store doesn't get there until it has the order data for a target, which Chapter 7 set at our gate of about 30 sales a month plus a few weeks of settled values. It's the same setting Chapter 10 covered for Performance Max, and the same advice applies: start with it off, or on with exclusions, and check where the clicks land.
Google Reports About 5% More Conversions at a Similar ROAS. Its Help Page Says to Wait Two Weeks Before Changing Anything.
Google puts the average gain from AI Max for Shopping at about 5% in conversions or conversion value with the return holding steady, based on its internal data from retail advertisers. That's Google's figure, and it's an average. We'd expect a store with sharp titles and a tight feed to see less, and a store with titles like "Mens Shoe SKU 123" to see more.
Google's reporting page sets the timeline for judging it: wait "at least 2 weeks after enabling" before making changes. Google's list of AI Max reports includes a product titles report. Read the titles Google wrote the way you'd read a new copywriter's work, title by title, before you read a single conversion number.
Are the titles accurate? Do they make claims you'd make? Would a shopper who clicked find exactly that on the page?
One reassurance carries over from earlier chapters. Google says you "generally don't need to add negative keywords to the Shopping campaign for terms you are already targeting in Search, as Exact match keywords in Search campaigns will take priority." The exact-match campaign from Chapter 5 generally keeps its searches.
Store owners tend to judge a new feature by the first week's clicks. In our view, that's the wrong number here: a broader title can attract clicks that don't turn into sales. Our way to judge it is an experiment: run AI Max on a copy of the campaign against the original for the same period, and compare purchases and conversion value, not clicks. If the test can't show a gain after the two weeks plus a full buying cycle, the campaign without it was doing fine.
AI Max Writes From Your Feed and Your Pages. Any Mistake in Either One Ends Up in Your Ads.
Everything AI Max writes comes from material the store already has: the feed, the website and the account's conversion data. That's why it comes last. If those inputs are wrong, AI Max can repeat the mistake in more ads and on more surfaces.
Before we switch it on for a store, these are done, in order:
- Purchase tracking with real order values, checked as primary (Chapter 1).
- Target ROAS running on real data in the Shopping campaign, above break-even (Chapter 7).
- A clean feed, with accurate titles, prices and availability (Chapter 14).
- Pages that agree with each other, so the returns page, shipping page and product page say the same thing (Chapter 8).
- Term exclusions set for every claim the store can't prove.
- An experiment, not a switch flipped on the whole account.
Chapter 9 covered the related setting at launch: Google can add AI-written content to Shopping ads unless you ask it not to. AI Max is the same idea with more reach. A store that has done the six steps is ready to let Google write. A store that hasn't is handing it the mistakes.
Next: Google Checks Your Price Before the Page's Scripts Run. A Price That Loads Later Can Get Products Disapproved.
AI Max can only be as good as the feed underneath it. Chapter 14 covers Merchant Center and the product feed: titles, prices and the price checks that can quietly take products out of Shopping.
Earlier chapters cover your brand campaign and Demand Gen and YouTube. Every chapter is on the guide's home page.
When we consider AI Max for a store through Google Ads management, the term exclusions and the experiment are set up before a single title is rewritten. Doing it yourself? Read your 20 best-selling product titles and list every claim word in them that you couldn't prove for every product.
