If you run Search campaigns with campaign-level broad match or standalone automatically created assets, Google is converting them to AI Max right now. The migration opened on September 1 and runs progressively through September 30. A campaign that looked untouched on September 2 can still convert on September 24.

Most of the coverage from August framed this as an opt-out story. That framing has expired. The window to opt out — by disabling the legacy features or manually enabling AI Max on your own schedule — closed before September 1. The useful question now is narrower and more practical: which cohort is each campaign in, when exactly did it flip, and what does that do to the analysis you're going to run in October?

The Timeline Google Actually Published

The mechanics came from Bob Hancock of the Google Ads API Team, published on the Google Ads Developer Blog on August 12, 2026 and reported in detail by PPC Land. The sequence is worth reading in order, because the enforcement ran ahead of the notice:

The block matters more than it sounds. Because it was enforced at the API level and in Editor, not just in the UI, any bulk sheet, script, or third-party tool whose campaign template quietly set campaign-level broad match started failing on August 3. If you build campaigns programmatically and haven't audited that path, it has been silently erroring for five weeks.

The Two Cohorts Land in Different Places

Google describes the migration as in-place conversion to "equivalent" AI Max settings, designed to minimize volatility. The equivalence holds better for one cohort than the other.

SettingBroad match cohortAutomatically created assets cohort
Search term matchingOnOn
Text customizationOffOn
Final URL expansionOffOff
Brand inclusion/exclusion listsPreserved

Read the first row against the column headers and the asymmetry jumps out. Automatically created assets governed creative generation — it wrote headlines and descriptions for you. Search term matching governs which queries your ad is eligible to serve against. A campaign that opted into automated copywriting and nothing else has now inherited automated query expansion by default.

That is a different category of setting than the one the advertiser originally consented to. It is documented, and Google's help pages carry the upgrade notice, so this isn't hidden. But "documented in a help article" and "the account team understood what they were agreeing to" are not the same standard, and the creative-to-targeting jump is the part most likely to have been skimmed.

Final URL expansion is off for both groups, which is the conservative call and the right one. It depends on text customization being active, so landing-page rerouting does not switch itself on at migration for either cohort. If you were bracing for traffic being sent to pages you never nominated, that isn't part of this event.

The Case That This Is a Non-Event

It's worth taking the deflationary read seriously, because for a lot of accounts it's correct.

For the broad match cohort, search term matching being on is close to a no-op. Those campaigns were already forcing every keyword to broad match — that is what the setting did. Query expansion was the existing behavior, not a new one. Google engineered the in-place mapping specifically so this cohort wouldn't lurch, and there's no strong reason to expect it will.

Even for the ACA cohort, the alarm assumes those campaigns were running tight phrase and exact keywords. Many weren't. If a campaign was already on broad match keywords with conversion-based Smart Bidding, search term matching adds incremental reach at the margin rather than opening a new door. And AI Max is a year old now, not a beta — the AI Brief steering controls that shipped alongside it exist precisely so advertisers can constrain matching and messaging in plain language.

So: for a meaningful share of accounts, the correct response to this migration is to note it and move on.

What Survives That Skepticism

One thing does, and it applies to every affected account regardless of cohort: your September data is structurally hard to read.

Because the migration rolls out progressively rather than on a single date, any account-level September-versus-August comparison blends pre-migration and post-migration traffic in proportions you don't know. That's not a small measurement caveat. It means the most obvious analysis anyone will run next month — "did AI Max help?" — returns a number that cannot answer the question.

Google evidently anticipated this, because the Developer Blog post shipped two new fields on the campaign resource for exactly this purpose:

campaign.aca_migration_date_time
campaign.broad_match_migration_date_time

Query those and you get a per-campaign timestamp of the moment its targeting behavior changed. That is the only precise marker available. Pull them, store them, and split every campaign's September into a before and an after around its own timestamp — not around September 1, and not around the calendar month.

This is the single highest-value action in this entire migration, and it has a deadline of its own: the fields tell you when things changed, but they don't reconstruct a baseline you failed to capture. If you have campaigns that haven't migrated yet, their clean pre-migration baseline is still available today and won't be in three weeks.

What to Do This Week

1. Pull the migration timestamps and store them

Query both fields across your affected campaigns and write the results somewhere durable. Do it now and repeat weekly through September 30, since campaigns are still converting. Without this, every performance conclusion you draw about AI Max this quarter rests on an unknown mixture.

2. Baseline the campaigns that haven't flipped yet

Any campaign with a null migration timestamp is a campaign whose clean "before" you can still record. Capture per-campaign cost, conversions, CPA, and — most importantly — the search terms report, at the campaign level rather than the account level.

3. Widen negative coverage before query matching widens

This is the one place where the ACA cohort deserves genuine attention. If a campaign inherited search term matching without ever having been a broad match campaign, its negative keyword list was built for a narrower matching regime. Review the incoming search terms report weekly for the first month post-timestamp, not monthly. Widening matching without widening negatives is the mechanism by which a migration month becomes a wasted spend month.

4. Audit your build automation

Anything that programmatically creates campaign-level broad match or legacy ACA configurations has been failing since August 3. Move campaign construction to standard Search campaigns with the AI Max settings bundle. Note also that once legacy entities transition into standard ad groups, attempts to modify them — as opposed to pausing or deleting them — return API errors. Treat those as expected, not as an outage.

5. Put February 2027 in the calendar now

Active legacy Dynamic Search Ads campaigns are excluded from the September migration and continue serving normally. Their window is February 1–28, 2027, and the destination has quietly narrowed since Google's June guidance: unmigrated DSA campaigns now upgrade to standard Search ad groups with AI Max enabled, where the earlier version had named Performance Max as a possible outcome. If you read our earlier write-up of the DSA upgrade, the dates in it have since moved — the schedule has now been revised twice.

6. Judge the outcome on incrementality, not the platform's framing

AI Max will produce attributed conversions. Whether total conversions rose is a separate question, and a progressive migration is a poor natural experiment. If the decision matters at your spend level, run a proper holdout rather than reading the before-and-after — the same discipline we've argued for on modeled conversions and incrementality.

The Pattern Worth Naming

Google closed creation on August 3, notified advertisers on August 5, and explained the mechanics on August 12. The enforcement arrived first, the notice second, the documentation third. That ordering is now common enough across ad platforms to plan around: the account is the source of truth about what is currently possible, and the blog post is the source of truth about why — and they do not arrive together.

The practical version of that lesson is unglamorous. Platform defaults change on the platform's schedule, and the leverage available to an advertiser has shifted from preventing the change to instrumenting it. This is the second Google default in as many months to arrive switched on — PMax AI video reframing did the same thing in early September, and that one at least had a live opt-out. Accounts that hold up under this pattern are the ones where a settings diff and a per-campaign timestamp are captured automatically, on a schedule, rather than assembled by hand after someone notices a number moved.

The migration is going to happen to your campaigns this month whether or not you look. The only part still under your control is whether, in October, you can tell what it did.

Sources: PPC Land — Google blocks new broad match settings before September 1 AI Max migration, Google Ads Developer Blog — Migrate Campaign-level Broad Match and Automatically Created Assets to AI Max, Relevant Audience — Google blocked new broad match settings on 3 August, Google — AI Max turns 1

Catch platform defaults the day they change

Ads Agents connects your AI agents to Google, Meta, and TikTok through MCP — so settings diffs, migration timestamps, and search term audits run on a schedule instead of after someone notices the number moved.

Get Started Free →