Google Help names seven elements of AI Max for Shopping campaigns. Three of them work with product data, and one of those three handles only English feeds as of September 22, 2026: text customization.
You're considering AI Max for a Czech catalog and want to know what you can use from it today. Each of those three elements takes its input from a different place.
I'll go through them one by one, following Google Help. Then I'll show four of my own feed enrichment measurements from Performance Max campaigns, with the caveats that belong to them.
What AI Max for Shopping campaigns is
Google Help describes AI Max for Shopping campaigns as a set of features that use Google AI to optimize Shopping campaigns and connect ads with customers running complex searches. The page carries a beta label.
It names seven elements: text customization, Final URL expansion, term exclusions, ad message restrictions, brand exclusions, URL exclusions and reporting changes. Below I cover the three where the product feed plays a role: text customization, Final URL expansion and reporting changes. Help puts four reports under the last one: product titles, search terms, landing pages and Final URL expansion assets.
Help does not list availability by country. You'll find out in Google Ads whether your account has the feature. Google has not published a launch date for Czechia.
Which features rely on the feed and what applies to a Czech feed
All three elements covered here use the product feed as their input, but each handles it differently.
| Feature | What it does according to Help | What it relies on | Czech feed as of September 22, 2026 |
|---|---|---|---|
| Text customization | Turns static product data into dynamic assets for long queries | Data in the feed | Help limits it to English feeds |
| Final URL expansion | Picks a more relevant landing page | The feed and the website content | Help lists no restriction |
| Reporting changes | Include four reports, among them product titles and search terms | Campaign data | Help lists no restriction |
According to Help, text customization converts static product data into dynamic assets that match long search queries. The English version of the page adds one more sentence about it: as of September 22, 2026, it is limited to English feeds. A Czech catalog will have to wait.
Final URL expansion replaces the final URL with a more relevant landing page based on the search query and the user's intent. So the content of your website has a say in where the visit lands.
Reporting changes are feedback for the feed. Look at the queries your product showed up on. Rewrite the title based on them. An illustrative example: you turn "Backpack 30 l" into "Hiking backpack 30 l with rain cover". The report doesn't tell you what such a change does to performance.
For the other two elements of this trio, Help mentions no language restriction. That is not a confirmation that they work with a Czech feed, though.
Feed enrichment in PMax: four of my own measurements
I don't have my own AI Max data yet. I have four feed enrichment measurements from client PMax accounts. In the first three cases I compared the enriched products with a control group from the same catalog. The fourth case compares the period before and after enrichment.
| Case | Products | Days | Result | Caveat |
|---|---|---|---|---|
| 1 (in Czech) | 48 | 22 | +719% conversion value and +260% clicks, both adjusted for the control group's trend | Raw clicks went from 480 to 1,582; without a control group that number says nothing |
| 2 (in Czech) | 138 | 22 | +144% clicks | Not reporting revenue yet, small conversion sample |
| 3 (in Czech) | 193 | 22 | +394% clicks | Mixed quality signal |
| 4 (in Czech) | 620 | 13 | +41% CTR in the successful subcategory | Average across all products only +8% |
In the first case, raw clicks grew from 480 to 1,582. Adjusted for the control group's trend, clicks grew +260% and conversion value +719%. Both numbers are adjusted, each measures a different metric. The raw number doesn't hold up on its own until a control group stands next to it. That's why every row has a caveat.
The fourth case ran for 13 days and shows why the result is calculated by catalog segment. One subcategory's click-through rate (CTR) jumped 41% against the period before enrichment. The average across all 620 products was only +8% in the same comparison.
Result
Case 4: CTR change against the period before enrichment
- Successful subcategory+41%
- Average across all 620 products+8%
The numbers come from Performance Max campaigns. They have nothing to do with AI Max or with AI shopping agents. All they say is that the data in the feed matters.
Before you turn on AI Max
Google gives one number for AI Max: advertisers who activate the feature typically see an average of 5% more conversions or conversion value at a similar CPA or ROAS. That is an average across advertisers, not a forecast for your account. Once you see the feature in your account, build your own test against a control group of products from the same catalog.
Until then, there's feed work that pays off either way. Go through the 10 most common disapproval reasons in Merchant Center. A disapproved product doesn't make it into an ad, however Google AI processes it.
Fill in the conversational attributes in Merchant Center and the values your feed is missing today. If you're waiting on a developer to change the export, you can add the missing values with a supplemental feed.
Summary
- AI Max is not one feature. Help describes seven elements under that name, and the three I cover each take their input from a different place.
- Text customization doesn't work with a Czech catalog yet. As of September 22, 2026, Help ties it to English feeds.
- Final URL expansion reaches beyond the feed. It picks the landing page based on the query, so the website content comes into play.
- Reporting changes feed data back to the feed. One of the four reports shows search terms, and those are the input for rewriting a product title.
- Five percent is Google's average. Don't read that number as a forecast for a single account.
- The four feed enrichment measurements come from PMax. You can't read an AI Max result from them, and every case has its own caveat.
- Your own test needs a control group. Without a comparable group of non-enriched products, you'll mostly measure the trend of the whole account.
AI Max changes how Google works with product data. What data you send it is still up to you.
Pick one category, fill in its missing attributes and leave an equally large group unchanged. The measurements above ran for 13 to 22 days. Don't judge clicks or CTR any sooner.
