For four years, the answer to “Performance Max is spending my money on things I cannot see” has been the same: move that part of the catalogue into Standard Shopping.
We have given that advice ourselves, with data behind it. In our Performance Max segmentation guide we documented an account where a product category was moved out of Performance Max and back into Standard Shopping specifically to regain control over search term exclusions. ROAS improved 66%.
Standard Shopping was the last campaign type where you could still see the queries, exclude what you did not want, and know roughly what you were buying.
Two changes this summer narrow that considerably.
On 24 July 2026, Google began rolling AI Max into Standard Shopping campaigns, bringing AI-powered query matching, ad text generation and landing page expansion to the campaign type that had least of it.
On 31 August 2026, Local Inventory Ads become default-on for every Shopping campaign, and the setting that let you turn them off is being removed entirely.
That second one has a hard deadline and a technical mechanic that will break automated setups. This guide covers both: what is actually confirmed, what Google’s performance numbers do and do not show, what independent testing found instead, and what we would do in the 32 days before the deadline.
Part 1: Local Inventory Ads, and the setting that is disappearing
Start here, because this is the one with a date and no opt-out.
What Google confirmed
On 20 July 2026, the Google Ads Developer Blog published a post by Dora Sun of the Google Ads API Team confirming that Local Inventory Ads will be enabled by default for all Shopping campaigns from 31 August 2026. Advertiser email notices had gone out four days earlier, on 16 July. Search Engine Land’s Anu Adegbola covered it the same evening.
This is a primary-source confirmation on Google’s own developer blog rather than a rumour, which matters given how much second-hand coverage this change has generated.
The mechanic
Three things happen on 31 August.
The “Local products” campaign setting is removed. It is replaced by the Inventory filter, which uses Channel = Local or Channel = Online.
The enable_local field on ShoppingSetting is effectively deprecated. The value is forced to true regardless of what you set.
On Google Ads API v25.1 and above, writing false to that field returns a hard error. Not a warning, not a silent ignore. Your call fails.
That last point is the one that will cause real problems. If you have scripts, a bid management platform, or any automated campaign creation logic that sets enable_local to false, that code will start failing on API v25.1 or later. Nobody gets an email about their scripts.
Who this actually affects
Be precise here, because a lot of the coverage has been alarmist.
No practical effect if: you sell exclusively online, have no local product feed in Merchant Center, and no verified store locations in your Business Profile. Local Inventory Ads cannot surface local inventory that does not exist. The default flips on and nothing changes.
Direct effect if: you have the LIA add-on enabled in Merchant Center and verified store locations. Local product feeds will start serving alongside your online Shopping ads on 31 August, which changes how budget is allocated and how conversions are attributed across every affected campaign.
Worth checking even if you think you are online-only: any retail distribution, pop-up shop data, or regional partner stores connected to your Merchant Center account. Local feeds attached to your account that you have forgotten about will start surfacing.
That last category is the one that catches people. Open Merchant Center and confirm what is actually attached before you assume this does not apply to you.
The two sanctioned ways to keep online and local separate
If you have been using enable_local = false to keep local inventory out of specific campaigns, you need a replacement before 31 August. Google’s sanctioned route is the Inventory filter:
Channel = Online on campaigns that should only serve online inventory. Channel = Local on campaigns that should only serve local inventory.
This is a genuine structural change rather than a settings tweak, particularly if your budget separation between online and in-store depended on the old flag. Work out which campaigns need which filter now, not on 30 August.
The strategic read
This aligns Standard Shopping with Performance Max for Retail, which has treated local serving as standard behaviour for some time.
That is the pattern worth noticing. Every one of these changes moves Standard Shopping closer to how Performance Max already behaves. Which is a reasonable description of the whole direction of travel.
For advertisers who never touched the setting, nothing visibly breaks. For advertisers who explicitly turned local serving off, the opt-out is gone.
Part 2: AI Max arrives in Standard Shopping
What it is
AI Max is a feature suite rather than a campaign type. It reached general availability for Search campaigns on 15 April 2026, and on 24 July 2026 Google began rolling it into Standard Shopping as a beta.
According to Google’s own documentation, AI Max for Shopping campaigns offers two things:
Text customization. Optimises your ad copy, including product titles, to make it more relevant across Shopping ads and to serve on newer AI-driven surfaces such as AI Overviews.
Final URL expansion. Unlocks text ads powered by your Merchant Center feed and dynamic landing pages, so you appear for more research-based searches.
You can enable both, or text customization only.
Why this matters more in Shopping than it did in Search
In Search, AI Max landed on a campaign type that had already been substantially automated. Broad match, Smart Bidding and automatically created assets meant most advertisers had already handed over the majority of the levers.
Shopping is different. Standard Shopping was where advertisers went precisely because it retained manual control: you could see the search terms, you could exclude aggressively, and your product titles were what you wrote rather than what Google generated.
Text customization changes the product title relationship. Final URL expansion changes the landing page relationship. Both are the specific controls that made Standard Shopping worth keeping.
That is not an argument against enabling it. It is an argument for testing it rather than assuming it behaves like AI Max for Search did.
Where the ads can now appear
Worth understanding what “AI-driven surfaces” actually means, since it is the upside of the change.
Ads are eligible above or below AI Overviews in all 200-plus markets where AI Overviews are available, and ads inside the AI Overview launched in the US. Ads in AI Mode, the conversational experience, began as a US experiment and are expanding through 2026.
The campaign types eligible for those placements are Performance Max, AI Max with search term matching, Shopping, and broad match.
So the genuine argument for AI Max is not the incremental conversion lift. It is placement eligibility on surfaces where an increasing share of commercial queries are being answered.
Part 3: The numbers, and who produced them
This section matters more than the feature descriptions, and it is where most coverage stops short.
What Google reports
AI Max for Shopping: an average 5% increase in conversions at a similar CPA or ROAS. Google’s documentation footnotes this as Google internal data, global, 2026, for retail advertisers.
AI Max for Search: an average 7% more conversions or conversion value at a similar CPA or ROAS, when using the full suite of search term matching, text customization and final URL expansion, compared to using search term matching alone.
A separate 14% figure appears for non-retail advertisers, from Google internal data in 2025.
Note the framing on the 7% number. It is not AI Max against no AI Max. It is the full feature suite against one feature of the suite. That is a narrower claim than it appears at first reading.
As Digital Applied put it when covering the same numbers, Google’s headline lift is a Google number: an in-platform average from its own Help Centre rather than an independent audit. Treat it as vendor-stated.
That is not an accusation of dishonesty. Vendors publish vendor data and label it, which Google does. It is a reason to weight it accordingly.
What independent testing found
Here is the part that changes the picture.
Reported results across 23 independent AI Max tests found that 84% of advertisers saw neutral or negative outcomes when measured at account level.
Two further findings from the same testing explain why:
Gains were smaller at account level than at campaign level. Much of the apparently new traffic was cannibalised from existing campaigns rather than representing net-new demand.
Accounts using all three AI Max features together outperformed those using single features, and text customization specifically lifted ad relevance and Quality Score in the majority of tests.
So the honest summary is not “AI Max does not work.” It is that AI Max can improve the campaign you measure while doing little for the account you own.
That distinction should be familiar if you have read anything else on this site. It is the same failure mode we documented when our own Performance Max benchmark turned out to be overstated by 5x because of how conversions were being counted. A number that improves inside one campaign is not the same as a business outcome.
A note on evidence quality, in both directions
We are not going to present the 84% figure as more solid than it is.
Google’s 5% is vendor-stated, clearly labelled, and drawn from a large sample the public cannot inspect. The independent 23-test figure comes from a much smaller sample, reported second-hand, and we have not been able to trace it to a published primary dataset with a full methodology.
Neither number should settle your decision. What they collectively tell you is that the effect is real but modest, that it varies enormously by account, and that where you measure it determines what you conclude.
Which means the only number that should decide this is one you generate yourself.
Part 4: When we would enable AI Max for Shopping, and when we would not
Our position is neither “turn it on” nor “avoid it.” It depends on four things.
Enable it if
Your feed is genuinely good. Text customization works from your product data. Brands with clean, complete, attribute-rich feeds get more from these tools than brands with sparse or outdated product data. If your titles are inconsistent and half your attributes are missing, AI Max amplifies that rather than fixing it.
You have enough conversion volume to read a test. If a campaign produces eight conversions a month, a 5% effect is invisible inside the noise. This is the same volume problem we found when seven of our ten Target CPA campaigns turned out to be below the threshold their bid strategy required.
Your conversion tracking measures real outcomes. If soft actions are inflating your conversion count, an incremental lift measured against that count means nothing.
You want the AI surface placements. This is the strongest argument. If a meaningful share of your category’s queries are being answered in AI Overviews, eligibility there is worth more than the headline percentage.
Hold off if
You moved that category into Standard Shopping deliberately, for control. If Standard Shopping is where you put products specifically to manage query matching tightly, enabling AI query matching undoes the reason you did that.
Your margins vary widely across the catalogue. Broader matching pulls in queries that convert at different values. Without segmentation and targets set from real economics, that shows up as volume that costs you margin. Our break-even ROAS calculator and ecommerce unit economics calculator are the place to start.
You cannot measure account-level impact. Given that cannibalisation is the main documented failure mode, if you can only see campaign-level numbers you will conclude it worked when it may not have.
You have another significant change running. Which, this month, most people do. More on that below.
Part 5: How to test it properly
The cannibalisation finding dictates the test design. If you measure only the campaign you enabled, you will get the answer Google’s numbers predict rather than the answer your account experienced.
Establish an account-level baseline first. Total conversions, total revenue, blended ROAS or cost per acquisition, across the whole account for 30 days. Not the campaign. The account.
Enable on one campaign with real conversion history. Not your smallest. Not your newest. Something with enough volume that a 5% effect is detectable.
Start with text customization only. Google supports enabling it alone, and it is the feature independent testing found most consistently positive. Final URL expansion is the bigger change and can wait until you have a read on the first.
Run it for at least two to four weeks. Long enough to clear the learning period and to see the query mix settle.
Watch the search terms report throughout. This is the whole point. AI query matching brings queries you did not choose, and the report is where you find out whether they are ones you want. Our Search Term Negative Keyword Analyzer will surface the waste faster than reading it manually.
Judge it at account level. Compare total account conversions and revenue to your baseline, not the enabled campaign to its own past. If the campaign is up 12% and the account is flat, you have moved conversions between campaigns and paid for the privilege.
Use the built-in controls. Brand settings, locations of interest, text guidelines, term exclusions and URL controls all exist. Turning AI Max on without configuring them is not a test, it is a surrender.
Part 6: The timing problem nobody is talking about
Look at what lands in a five-week window.
17 August: budget-limited campaigns on target-based bid strategies begin optimising toward their stated target rather than potentially overperforming it. Shopping is included.
31 August: Local Inventory Ads default-on for all Shopping campaigns, enable_local deprecated.
September: campaigns using Automatically Created Assets and the campaign-level broad match setting begin being auto-upgraded to AI Max.
Ongoing: the AI Max for Shopping beta rollout, and the Display to Demand Gen migration running in parallel.
If you enable AI Max on 20 August, your Shopping campaigns will simultaneously be adjusting to new bidding behaviour, absorbing local inventory serving, and running broader query matching. When performance moves, you will have no way to attribute it to any of the three.
Our recommendation on sequencing:
Deal with the Local Inventory Ads deadline first, because it has no opt-out and a technical failure mode. Get your Inventory filters set and your scripts fixed.
Let the 17 August bidding change settle. Google itself advises caution with forecasts between 17 and 31 August.
Then test AI Max in September, from a stable baseline, as a deliberate experiment with its own read.
Doing all three at once is not efficient. It is the fastest way to learn nothing from any of them.
Part 7: A 32-day checklist
This week: the Local Inventory Ads audit
- Open Merchant Center and confirm whether the LIA add-on is enabled
- Check for verified store locations in your Business Profile
- Identify any retail distribution, pop-up or partner store feeds attached to the account that you had forgotten about
- List every campaign currently using
enable_local = false - Search your scripts, rules and any third-party platform for code that writes to that field
Next week: implement the replacement
- Set Inventory filters, Channel = Online or Channel = Local, on campaigns that need separation
- Update or remove any script writing
enable_local - Confirm your API version and whether v25.1 or later applies to you
- Audit local feed freshness. Stale local inventory data will now serve by default, which means pointing people at stock you may not have.
Week three: baseline everything
- Record 30 days of account-level performance before any of these changes land
- Note branded impression share specifically, since it is a useful early indicator of query mix shifting
- Confirm your conversion tracking is measuring real outcomes. The 9-Pillar Scorecard covers this alongside structure and unit economics.
Weeks four and five: hold
- Let 17 August and 31 August pass
- Monitor, do not react
- Watch impression share on branded terms through early September
September: test AI Max deliberately
- One campaign, text customization only, controls configured, account-level measurement, two to four weeks
Part 8: The pattern underneath all of this
Step back and the individual changes matter less than the direction.
Manual bidding options are disappearing. Campaign types are consolidating. The controls that let advertisers steer are being replaced by controls that let advertisers constrain, which is a different and smaller thing.
Standard Shopping is now getting AI query matching and generated ad text. Display has folded into Demand Gen. Local Services Ads are migrating into Performance Max, beginning in August 2026 with US home and storefront services including plumbing, HVAC, electrical, roofing and pest control. Dynamic Search Ads have upgraded to AI Max. Campaigns using automatically created assets and campaign-level broad match are being auto-upgraded from September.
The strategic consequence is that your inputs matter more than your levers.
Feed quality, creative quality, conversion tracking integrity, unit economics and campaign structure are the things you still fully control, and they are increasingly the only things that differentiate one advertiser from another inside the same automated system.
That is not a comfortable message if your value came from knowing which settings to change. It is a good message if you have been doing the unglamorous work. Everything we have published this year points the same way: conversion tracking and unit economics come before bidding, and structure decides whether automation has enough data to function.
The advertisers who will do well through this are the ones providing better inputs, not the ones trying to outmanoeuvre the algorithm.
Frequently asked questions
What is AI Max for Shopping campaigns? A beta feature suite that brings AI-powered query matching, text customization and Final URL expansion to Standard Shopping campaigns. Text customization optimises ad copy including product titles for relevance across Shopping ads and AI-driven surfaces. Final URL expansion unlocks text ads powered by your Merchant Center feed and dynamic landing pages. You can enable both or text customization alone.
Does AI Max actually improve performance? Google reports an average 5% conversion increase at similar CPA or ROAS for Shopping, and 7% for Search when using the full feature suite versus search term matching alone. Both are Google internal figures. Reported independent testing across 23 accounts found 84% saw neutral or negative results when measured at account level, with much of the apparent gain coming from traffic cannibalised out of existing campaigns. Test it yourself and measure at account level.
When do Local Inventory Ads become default? 31 August 2026, confirmed by Google on its Ads Developer Blog on 20 July 2026, following advertiser email notices on 16 July.
Can I opt out of Local Inventory Ads? Not through the old setting. The “Local products” campaign setting is being removed and the enable_local API field is being forced to true. Use the Inventory filter with Channel = Online or Channel = Local to control which campaigns serve which inventory.
What happens to my scripts? Any code writing enable_local = false will stop working. On Google Ads API v25.1 and above, that call returns a hard error rather than being ignored. Audit your scripts and any third-party platform before 31 August.
I sell only online. Does this affect me? Probably not. Local Inventory Ads can only surface inventory that exists in a local feed. If you have no LIA add-on in Merchant Center and no verified store locations, the default flipping on changes nothing. Check for forgotten partner or pop-up feeds attached to the account before concluding that.
Should I enable AI Max before or after 31 August? After, and ideally in September. Between 17 and 31 August your Shopping campaigns are already absorbing a bidding behaviour change and a local inventory change. Adding broader query matching on top means you will not be able to attribute any performance movement to any single cause.
Does AI Max get turned on automatically? For Shopping it is currently an opt-in beta. Separately, Google has said campaigns using Automatically Created Assets and the campaign-level broad match setting will continue to be auto-upgraded starting September 2026, so some accounts will be moved without action.
How do I test AI Max properly? Baseline at account level for 30 days first. Enable on one campaign with real conversion history. Start with text customization only. Run two to four weeks. Watch the search terms report throughout. Judge the account, not the campaign, because cannibalisation is the documented failure mode.
Is Standard Shopping still worth using instead of Performance Max? Often yes, and we have published a case where moving a category out of Performance Max into Standard Shopping produced a 66% ROAS improvement through better query control. AI Max narrows the gap but does not close it, and Standard Shopping still gives you visibility Performance Max does not. Our guide to Performance Max segmentation covers when each is appropriate.
What should I fix first if my feed is weak? The feed, before anything else. Text customization works from your product data, and the tools reward complete, attribute-rich feeds. Titles, images and price carry most of the weight in Shopping performance regardless of which AI features are enabled.
About The Author
