Google AI Max for Shopping: What Sellers Should Keep Control Of

If you run Standard Shopping campaigns for an ecommerce brand, two things changed in your account in 2026, and one of them happened without your input.

The first is AI Max for Shopping, announced by Google on April 30, 2026 and spotted in live accounts from late July. It is an opt-in beta that adds three behaviours to campaigns you already run: Google can rewrite your product titles at serve time, send a click to a landing page other than the one in your feed, and choose whether a Shopping ad or a text ad serves for a given query.

The second is not an AI Max feature, but it lands on the same settings page. On August 31, 2026 Google enabled Local Inventory Ads by default for every Shopping campaign and removed the "Local products" opt-out. If you have any store, retail partner, or regional pickup data connected to Merchant Center, local offers can now serve through campaigns that were online-only last month.

Neither change touches your product groups, your bidding strategy, or your negative keywords. What both change is where the decision gets made. Fewer decisions sit in a settings box, more sit in a report someone has to read every week. This post covers which controls still stop something, which reports you now have to work, and what the extra reach realistically costs.

Quick answer

• What it is. An opt-in beta that layers three features onto Standard Shopping: text customization (Google rewrites product titles per query), Final URL expansion (Google picks the landing page and unlocks feed-powered text ads), and format selection (Shopping ad or text ad).
• What you keep. Product groups, bidding strategy, and negative keywords carry over unchanged. The upgrade is a one-click toggle, and you can reverse it.
• What it gets you. Reach into research-stage queries your feed could never trigger, because a shopper searching "best shoes for standing all day" never types your product title.
• What it costs. Google reports a typical 5% lift in conversions or conversion value at similar CPA or ROAS. Independent Search-side data shows a much wider spread, with CPA rising alongside conversion value. Budget for incremental demand costing more than the demand you already had.
• What to switch on first. Text guidelines, brand settings, URL exclusions, and the inventory filter. These are the four controls that actually stop something from happening.
• What to read weekly. The landing pages report, the expanded final URL assets tab, and the search terms report. These record rather than prevent, so they only pay back if someone reads them and acts.
• What already changed without you. Local Inventory Ads have been on by default since August 31, 2026. Channel control moved from a checkbox to the inventory filter.

What AI Max changes inside a Standard Shopping campaign

Google's framing is that Shopping ads are static and shoppers are not. Someone searching "best high-quality clothes for lounging" never types your product title, so a feed-triggered ad does not enter the auction. AI Max for Shopping is Google's answer to that gap, built on the Merchant Center feed rather than on keywords.

Text customization rewrites your product titles at serve time

With text customization on, Google generates product titles from your feed attributes and serves them in place of the title you uploaded. Google's documentation gives the shape of it: "Mens Shoe SKU 123" can serve as "Men's Waterproof Hiking Boot" when the query justifies it.

Two guardrails are worth knowing. Google says generated titles are checked against your actual product data before they can serve, and that a customized title only serves when it is predicted to outperform the title you uploaded. Your original title serves otherwise. That is a meaningful constraint, and it is also a prediction rather than a promise, which is why the product titles report exists.

If you have spent real time on feed hygiene, this is the setting that will feel strangest. Text customization leaves your feed file alone. What it decides, per query, is whether your title or its title has a better chance, and the record of that choice sits in reporting rather than in your feed. The same discipline you would apply to product content optimization still applies, just after the fact instead of before.

Final URL expansion picks the landing page and brings text ads with it

Final URL expansion stops sending every click to the product URL in your feed. Google crawls your site, identifies other commercial pages such as category pages, new arrivals, or editorial pages, and routes query-relevant traffic there. Google states the expansion only sends traffic to URLs that are query relevant and themed to the ad group.

One dependency catches people out: Final URL expansion requires text customization to be turned on, so the ad copy matches the destination. You cannot take the landing page expansion without the title rewriting. You can take the title rewriting on its own.

Expansion is also what unlocks text ads powered by your feed, which is how a Shopping campaign starts competing for research-stage queries it never touched before. That is the incremental reach on offer, and it is also the part most likely to send a shopper to a page nobody on your team chose.

Format selection decides between a Shopping ad and a text ad

The third piece is quieter. Google picks the format, Shopping or text, based on what it reads as most relevant to the query. There is no separate report for the decision itself, only the performance of what served.

The 2026 Google Shopping dates that changed your account

The Shopping timeline moved fast this year. One of the entries below is not an AI Max feature at all, but it lands on the same settings page and it is the only one that took effect whether you opted in or not.

The August 31 change is the one to check today, and it is worth a few minutes even if you have never opened the Google Ads API. Local Inventory Ads are the Shopping listings that show a product as available in a nearby store rather than only online. Until August 31 you could keep them out of a campaign by unticking a box called "Local products." Now you cannot. Google has enabled Local Inventory Ads by default in Shopping campaigns, aligning them with Performance Max for Retail.

If you manage campaigns through the Google Ads API, the enable_local field no longer has any effect: the API overrides it to true, and on version 25.1 and later, setting it to false returns a ContextError.OPERATION_NOT_PERMITTED_FOR_CONTEXT. If you previously used that field to keep local offers out of a specific campaign, the replacement is a listing scope with product_channel set to ONLINE. In the campaign settings UI, the equivalent is the inventory filter, which lets you separate budgets by channel. Performance Max and Demand Gen are unaffected.

For an online-only retailer with no local feed, nothing changes. If you have any retail distribution, store partners, or regional pickup data connected to Merchant Center, local offers can now serve through campaigns that were previously online only, and you have to filter rather than untick.

What Google's 5% lift figure actually measures

Google publishes one performance figure for this feature. Its help documentation states that advertisers who activate AI Max for Shopping typically see 5% more conversions or conversion value at a similar CPA or ROAS, attributed to Google internal data, global, 2026, for retail advertisers. Directionally that is a modest, believable claim for an expansion layer, and it is more restrained than the numbers attached to AI Max for Search.

It is also an average, which is where a second data point earns its place. Smarter Ecommerce analysed more than 250 Search campaigns running AI Max and found a median conversion value lift of 13% alongside a median CPA increase of 16%, with ROAS outcomes ranging from 42% above baseline to 35% below, and only 22% of campaigns landing near their original ROAS target. That study covers Search rather than Shopping, so read it as a pattern rather than a forecast. The pattern is that incremental demand costs more than the demand you already had, and that the spread across accounts is wider than any single average suggests.

Plan the test accordingly. Google recommends waiting at least two weeks after enabling AI Max for Shopping before making changes, so the campaign has room to ramp. Budget the two weeks, and decide before you start what result would make you turn it back off.

Off switch or audit trail: which AI Max controls stop something

Final URL expansion has an off switch. The landing pages report does not. That difference is what decides which controls on the settings page are worth defending, because every one of them is one of two kinds.

An off switch stops something from happening. You untick it and the behaviour ends. Final URL expansion is an off switch. So is text customization. So was the "Local products" setting, right up until August 31.

An audit trail records what happened so you can react afterwards. The landing pages report, where Google-chosen destinations carry a "Selected by" value of "Automatically selected", is an audit trail. So is the expanded final URL assets tab, which shows the asset and URL combinations that served and lets you remove the ones that do not fit. So is the search terms report.

Both are worth having. They cost different things. An off switch costs you the reach behind it and nothing else. An audit trail costs you a recurring hour, every week, forever, and it only pays if someone reads it and acts on it. The failure mode with automation is rarely a missing control. It is usually that the control was a report, and nobody had the time to read the report.

That framing explains the August 31 change. Local inventory did not become a worse idea. It moved from the first category to the second. Anything you want excluded now has to be filtered and monitored rather than switched off, and that is a standing job rather than a one-time decision. The same shift is what makes choosing between automated and rule-based logic a live question rather than a settled one.

The AI Max controls to set before you flip the toggle

Google's best-practice guidance is to turn on both text customization and Final URL expansion for maximum performance. That is sound advice for reach, and it is not the same as advice for control. A sequence that gives you both:

• Turn on text customization first, alone. It is the smaller change, it does not touch your landing pages, and it produces the product titles report you need in order to judge whether Google's rewrites read like your brand.
• Write text guidelines before you generate a single title. This is the cheapest control on the page. It tells the system what your ads should and should not say, and it costs one sitting.
• Set brand settings deliberately. Brand exclusions are a campaign-level control and brand inclusions work at campaign or ad group level. If you have competitor-name policies or trademark commitments, encode them here rather than discovering them in a search terms export.
• Add Final URL expansion as a separate decision, with URL exclusions ready. Pull the list of pages you never want an ad pointing at, checkout flows, out-of-stock category pages, legacy campaign landers, and load them as exclusions on day one rather than after the first bad week.
• Set your inventory filter now. With the local products opt-out gone, the inventory filter is how you separate online and local serving and stop one budget from funding both.
• Keep the search terms report on the weekly list. Google's guidance is not to over-optimize with negatives for terms already covered by exact match keywords in your Search campaigns, since those take priority. Reading the report is still how you find out what conversational query matching matched you to in practice.

Standard Shopping is a channel many brands run alongside marketplace advertising rather than instead of it, and Google Shopping's mechanics reward attention that a busy Amazon-first team rarely has spare. That is worth being honest about when you decide how much of this list you will genuinely do.

What you can set, see, and stop, by campaign type

That last row is the uncomfortable one. Turning AI Max on gives you more reach and more weekly work at the same time, and the weekly work is the part that quietly gets dropped in month two.

Where the weekly AI Max review breaks down

A careful operator can run this well. Week one you read the landing pages report, flag two category pages you do not want, add them as URL exclusions. Week two you scan the expanded final URL assets tab and remove three generated headlines that read wrong for the brand. That work is real, and a general-purpose AI tool will help you do it faster: paste an export in, ask for the outliers, get a usable shortlist in a minute.

Analysis is the easy part. The hard part is the fourth week, on the eleventh account, when the person who defined "a landing page we do not want an ad pointing at" is on holiday and the standard exists only in their head. The reports still generate. The judgment that turns a report into a decision does not scale by being written down once in a shared doc, and asking a chat tool the same question every week returns a slightly different answer each time, which makes the weeks incomparable.

That gap is what Qore addresses. You take a review you currently do by hand, describe it once in your own words, and turn it into logic you can see and lock, so the same standard runs every week and returns the same output from the same inputs. For AI Max for Shopping that looks like a scheduled review that reads the landing pages and expanded final URL reports against your own criteria and flags the destinations and generated assets that fail them.

How much rein it gets is a setting you tune. Start with every exclusion gated on your approval, watch it make the same call correctly for a month, then tighten the rules until you are comfortable letting it apply the clear-cut cases on its own and escalate only the ambiguous ones. A skill can act, not just recommend, once you have made it conservative enough to trust. The parallel on the marketplace side is the same shape as a repeatable PPC account audit workflow: the value is in running the identical review every time, not in running a clever one once.

Two honest limits. Qore codifies a standard, it does not invent one: if nobody has decided which landing pages are acceptable, there is nothing to encode yet. And channel coverage is not channel equivalence. Trellis goes deeper on Amazon than on Google, so treat Google Shopping here as a workflow you can run consistently rather than a channel with feature parity to the marketplace side. Anyone running advertising across a wide catalog and several channels already knows the difference.

What to do this month

If you run Standard Shopping, do three things before the holiday period locks your account down.

Check your inventory filter first, because that change already took effect on August 31 and it is the only one on this list that happened without you. Then, if you want to test AI Max, turn on text customization by itself, write text guidelines before the first generated title serves, and give it two weeks before you judge anything. Add Final URL expansion as a second, separate test with your URL exclusions already loaded, and treat the landing pages report as the thing that decides whether it stays.

The bottom line for ecommerce teams

• The Local Inventory Ads change is not optional and already happened. Check your inventory filter this week. If you have store or pickup data in Merchant Center, an online-only budget may now be funding local serving.
• AI Max is optional, reversible, and worth a controlled test. Expect more reach at a higher cost per conversion, not free incremental revenue.
• Turn on the four controls that actually stop something before you flip the toggle: text guidelines, brand settings, URL exclusions, inventory filter.
• Be realistic about the weekly reports. Three reports, every week, across every account. If nobody owns that hour, the reach you gained will cost more than it returns by month three.
• Write down what a bad landing page looks like for your business. That sentence is worth more than any toggle on the page, and it is the same sentence that makes automation safe to run once you have it.