You did not flip AI Max on. September still changed your Search settings.
Google is progressive-migrating Search campaigns that still use campaign-level broad match or standalone Automatically Created Assets into AI Max between September 1 and September 30, 2026. Bob Hancock of the Google Ads API Team laid out the mechanics on the Google Ads Developer Blog on August 12, 2026: cohort-specific defaults, two new migration timestamp fields, a creation block already in force since August 3, and Dynamic Search Ads parked until February 2027. Search Engine Land reported the advertiser email on August 5. Tech Help Canada, writing on September 19 while the window was open, called it what it is: a progressive rollout through September 30, not a single-day flip.
If your mental model is “we never enabled AI Max, so we are fine,” you are auditing the wrong object. The September job is not a yes/no product opinion. It is a settings audit: search term matching, text customization, Final URL expansion, negatives, brand lists, and generated assets, campaign by campaign, with migration timestamps recorded so you can separate seasonal noise from the change.
What actually changed
Three facts matter. Everything else is noise.
Fact one: two cohorts, not one feature flip. Per the August 12 developer post (as detailed by PPC Land the same day and verified in Choice OMG’s August 26 dossier), campaigns on campaign-level broad match migrate in place with search term matching on, text customization off, and Final URL expansion off. Brand inclusions and exclusions are preserved. Campaigns on standalone Automatically Created Assets migrate with search term matching on, text customization on, and Final URL expansion off. Same umbrella. Different defaults. An account can show AI Max active without every AI Max sub-feature being active.
Fact two: creation already died on August 3. New campaign-level broad match configurations and legacy ACA are blocked across the UI, Google Ads Editor, and all API versions as of August 3, 2026. The population that can still auto-migrate is fixed. You cannot accidentally grow it. You can still inherit it if those settings were already on.
Fact three: DSA is not this month. Brandon Ervin’s April 15, 2026 product post originally put DSA, ACA, and campaign-level broad on a September track. The June 11 update moved DSA sunset and auto-upgrade to February 2027 while keeping ACA and campaign-level broad on September. The August 12 developer post restates the DSA path: in-account banners in September 2026, a January 15, 2027 reminder, automigration February 1 to 28, 2027, into standard Search ad groups with AI Max enabled. If someone on your team is panicking about DSA this week, they have the wrong calendar.
Alphabet’s Q2 2026 earnings call (July 22) claimed half a million advertisers already using AI Max. That is an adoption count, not the size of the September-eligible population. Choice OMG is explicit: no credible public denominator exists for how many Search campaigns still sit on those two legacy settings. Do not invent one for a leadership slide.
The mechanism: defaults landed without a click
AI Max is not a new campaign type. Google’s own framing treats it as an optimization layer on existing Search campaigns: search term matching (query expansion via broad match and keywordless tech), text customization (generated headlines and descriptions), and Final URL expansion (page selection). Migration stamps the umbrella on and maps your legacy setting into a default bundle.
That is why “I never clicked AI Max” is a useless sentence. You may have clicked campaign-level broad match years ago because Smart Bidding wanted it. You may have left ACA on because a recommendation looked harmless. The September job converts that past choice into today’s AI Max panel with cohort defaults, on a progressive schedule you do not control calendar-day by calendar-day.
The API side makes the progressivity measurable. Hancock’s post adds campaign.aca_migration_date_time and campaign.broad_match_migration_date_time. Query them. A campaign quiet on September 2 can still change on September 24. Month-level before/after without those timestamps blends pre and post traffic into mush. Common Thread Collective (August 6) also stressed the operator distinction operators still miss: this targets the campaign-level broad match toggle, not every ad-group broad keyword you deliberately wrote. Ad-group broad without the campaign setting is not the September trigger.
Controls survive under current documentation. Choice OMG’s re-check of Google help (accessed August 25, 2026) is clear: after conversion you can still toggle the AI Max umbrella and the sub-settings. Search term matching can be disabled per ad group. Text customization can be turned off (generated assets then report as Removed and stop serving). Final URL expansion cannot stay on without text customization. “Permanent forever opt-out” is not what Google contracted. “You can change it under today’s UI” is.
What operators get wrong
Mistake one: treating AI Max as a feature you never clicked. Leadership asks “are we on AI Max?” Media answers “no.” Meanwhile campaign settings already show the umbrella with search term matching on because the campaign sat on legacy broad. The right question is “which campaigns migrated, on which date, with which defaults, and do those defaults match intent?”
Mistake two: conflating campaign-level broad with ad-group broad keywords. Common Thread Collective made this the ecommerce warning on August 6. If you never flipped the campaign-level setting, ad-group broad keywords alone do not put you in the September cohort. Wasting the week rewriting every broad keyword list while ignoring the campaign toggle is theater.
Mistake three: accepting “mirrors legacy” as “identical.” Google’s language is continuity and reduced volatility. Continuity is not parity. ACA cohorts inherit search term matching they did not previously have as a named AI Max control. That is query expansion arriving with creative automation. Broad cohorts inherit the umbrella and search term matching on while creative stays off. If you do not open the AI Max panel, you will not see the asymmetry.
Mistake four: skipping generated asset review on ACA cohorts. Text customization on means Google-created headlines and descriptions can serve with your RSAs. Brad Geddes’s Search Engine Land write-up of Adalysis-guided tests (July 28, 2026), summarized carefully in the Choice dossier, found directionally mixed results: long-tail help in neglected campaigns, mature and B2B pain when generated copy pulled the wrong audience, and roughly 19% of non-B2B generated assets removed early in that small sample. That is not a market average. It is a reason to inspect assets before you call September “fine.”
Mistake five: freezing negatives and brand lists because “we will wait for the migration to finish.” Search term matching expands eligibility. Tech Help Canada (September 19) puts search terms, negatives, brand controls, and assets on the same checklist as the AI Max columns in campaign list view. Waiting until October to mine queries is how Q4 learns on garbage.
Mistake six: stuffing DSA panic into the September ticket. Wrong window. Wrong deliverable. Wrong upgrade destination story if you are still quoting the April post without the June correction. Park DSA as a separate 2027 project: inventory dynamic ad groups, page feeds, part-number use cases, URL rules. Do not let it steal this week’s audit hours.
Mistake seven: reading Google’s 7% as your account math. Ervin’s April/June post cites an average 7% more conversions or conversion value at similar CPA/ROAS for the full suite versus search term matching alone, Google internal non-Retail data. Independent evidence is mixed. Smarter Ecommerce (March 23, 2026) reported median +13% conversion value, median +16% CPA, and median ROAS difference of 0% across more than 250 retail Search campaigns. Lunio (updated August 14, 2026) reported rising invalid-traffic rates on AI Max versus standard Search in its observational retail Search cohort. Location3 (March 18, 2026) reported conversion lifts in five franchise experiments with limited external validity. Cite those as mixed third-party signals, not as a reason to skip the audit or to torch every AI Max switch without looking.
Second-order effects
Media and SEM ownership collide with creative compliance. ACA cohorts put generated copy into the auction without a weekly creative review unless you staff one. Regulated categories, local services, and tight B2B claims cannot treat text customization as a free performance lever.
Agency scopes split again. You need a migration cell that owns timestamps, cohort labeling, and the three-setting matrix, separate from the person who “owns Search ROAS.” Choice OMG’s field guide is blunt: decide campaign by campaign among prevent (turn off legacy before migration), manual AI Max with chosen settings, or accept cohort defaults. Account-wide dogma is how brand and catalog campaigns get the same wrong bundle.
Reporting politics sharpen. Match type “AI Max,” Source columns, and AI Max expanded-match totals will disagree with last year’s keyword story. Low-volume queries still hide under privacy thresholds. If your Q3 readout cannot show migration dates beside performance deltas, you will argue seasonality forever.
Tooling and scripts get a silent break. Legacy creation already fails. New API versions after September 1 remove legacy entities; older versions linger toward a roughly September 2027 sunset. Agencies still writing campaign-level broad via API are already broken. That is a vendor ticket, not a media opinion.
Q4 timing is the real pressure. The window closes weeks before peak retail. Defaults that look “fine” on a quiet Tuesday become expensive when budgets concentrate. Alphabet’s half-million AI Max advertiser claim is the competitive backdrop, not your success metric.
What to do this week
- Build the migration roster. Campaign list view with AI Max columns and filters (Tech Help Canada, September 19). Export name, status, cohort trigger (campaign-level broad vs ACA vs neither), current search term matching / text customization / Final URL expansion, brand lists, and the two API migration timestamps where available. Anything still unmigrated gets a watch date through September 30.
- Separate the three decisions from the three settings. For each affected campaign: keep umbrella on or off; keep search term matching on or disable per ad group; keep text customization on or off (knowing Final URL expansion depends on it). Write the intended state next to the observed state. Close the gap or document why you accept it.
- Harden negatives and brand controls before you debate ROAS. Account and campaign negatives for brand-adjacent junk, competitor leakage you refuse to buy, and category mismatches. Confirm brand inclusions/exclusions survived on broad-cohort migrations. This is cheaper than a post-hoc “AI Max bad” narrative.
- Run an asset and landing-page pass on ACA cohorts. Pull Google-created text. Remove claims you will not stand behind. Confirm Final URL expansion is still off unless you explicitly want page selection. If any campaign has Final URL on, review URL exclusions before spend wanders.
- Instrument the before/after with timestamps, not vibes. Two-week post-migration windows per Hancock’s progressive design. Watch search terms labeled AI Max, conversion quality, branded and competitor leakage, and account-level revenue, not only treatment-campaign ROAS (Choice dossier synthesis of cannibalization risk).
- Park DSA on a February 2027 track. One owner, one inventory sheet, no September panic. Ervin June 11 and Hancock August 12 agree on the revised calendar.
- Brief GTM and agencies in operator language. The story for leadership is not “Google forced AI on us.” It is “eligible Search campaigns are mapping into AI Max with cohort defaults through September 30; we are auditing settings, not toggling a philosophy.” Marketers and agencies who ship that brief this week stop the Slack rumor mill from inventing a DSA emergency.
Sharp close
The upgrade already happened for someone on your roster, or it will before the month ends. Clicking nothing was never a strategy. Open the settings. Read the defaults. Own the matrix. September is an audit month. Treat it like a toggle and you will discover your search term matching policy in the query report after the spend has already learned.
