Google Is Migrating Your Search Campaigns This Month. Set the Baseline Before It Does.

On September 1, Google started moving campaign-level Broad Match and legacy Automatically Created Assets campaigns into AI Max, and the migration keeps rolling through the month, account by account, without a confirmation step. Microsoft made its own AI Max generally available the same week and turned it on by default for every new Search campaign. On October 1, Max CPC disappears from new non-portfolio campaigns. Three platform decisions, one direction, and a calendar that does not wait for your Q4 planning meeting.

None of this is a scandal. Both companies have been telling advertisers for two years that the future of search buying is a single AI layer that expands queries, writes assets, and picks landing pages on its own. The problem with September is subtler than a bad product. Every campaign that gets migrated this month will report its first AI Max results during the strongest seasonal weeks of the year, and the platform will present those results as evidence. Without a baseline captured before the switch, you cannot tell whether AI Max improved the account or whether October happened to it.

So the work this week is measurement. Pull the list of campaigns in the migration path, which for most accounts means anything still running campaign-level Broad Match or an old ACA structure. For each one, freeze a thirty-day pre-migration snapshot: spend, conversions, cost per conversion, conversion rate, impression share, and the search term report exported in full, because the search term report is the document that changes most under AI Max and the one you will want to compare against in November. Store it somewhere the platform cannot overwrite. Google's own reporting will show you AI Max performance against nothing in particular, and nothing in particular always loses.

Then decide what you are willing to lose control of. AI Max is a bundle of behaviors, and the bundle is where the damage hides. Query expansion lets the system match your ads to searches you never bid on, which is fine in a category with clean intent and expensive in one where adjacent queries mean adjacent products. Asset generation writes headlines in your name, which regulated advertisers cannot accept without a review workflow that most accounts do not have. Landing page selection sends traffic where the model thinks it converts, which sometimes means the page your compliance team retired in March. Each of these has an exclusion or a control, and each control has to be set before the migration, because the migration is the moment defaults get inherited.

Microsoft, to its credit, shipped the reporting that Google left out. The Microsoft version ships an "AI optimized" match type label in reports, visibility into AI-generated assets in the asset report, a search term landing page report, and term exclusions. If you have spent two years arguing that opacity is a performance risk in Performance Max, Microsoft just agreed with you in product form, and Bing's share of the AI retrieval layer makes that account worth more attention than its click volume suggests.

The measurement design that settles the question is not complicated. Split the account. Leave a matched set of campaigns unmigrated for as long as the platform allows, migrate the rest, and read the two groups against each other on the same calendar instead of reading the migrated group against its own past. Seasonality hits both halves equally, so the difference between them is the product. You do not need a data science team for this; you need two campaign groups and the discipline to leave one alone. Where a holdout is impossible because Google has already moved everything, fall back to the frozen snapshot and compare conversion rate and cost per conversion rather than volume, because a rising click-through rate will happily hide a falling conversion rate inside a migration narrative that says everything improved.

What I expect, from the accounts I have already watched cross over, is a familiar shape. Volume goes up. Cost per conversion drifts up more slowly, so the first report looks like growth. The search term report fills with queries that a human would have negated on sight, and by the third month the account is paying for reach it never wanted at a price it never set. Advertisers who caught it in week two spent one afternoon on exclusions and kept the upside. Advertisers who read the November report cold could not reconstruct what changed, because the platform had already overwritten the record.

The migration is happening either way. The baseline is the only part of this you still control, and it expires the day your campaign moves. Export it this week.