The Google Ads auto-upgrade most accounts don't know they're on

If your account uses Automatically Created Assets or the campaign-level broad match setting, you are on Google's September 2026 auto-upgrade list. It does not matter whether you have ever run a Dynamic Search Ads campaign. Most of the commentary is watching the DSA date and will get caught by the other one.

Here is the full picture, because the sequencing matters more than the feature.

What AI Max actually is

AI Max for Search is a campaign-level setting, not a campaign type. It bundles three things:

1. 1.Search term matching. The campaign can match queries beyond your keyword set, using the page content and your existing keywords as a guide.
2. 2.Text customisation. Headlines and descriptions get rewritten per query, drawing on your landing pages and existing assets.
3. 3.Final URL expansion. Google picks a landing page from your site rather than only using the one you nominated.

Google's published result for the full suite is around 7% more conversions or conversion value at similar CPA or ROAS, measured against running search term matching on its own. Take that as directional. It is aggregate data across a huge base of advertisers, and your account is not the average advertiser.

The dates

Voluntary upgrade tooling is available now. From September 2026, auto-upgrades begin for campaigns using Automatically Created Assets or the campaign-level broad match setting, and you can no longer create new DSA campaigns. In February 2027, the remaining DSA campaigns get upgraded whether you act or not. Google pushed that final date back from an earlier September 2026 deadline, which is why some calendars are wrong.

The trap is the first date. Automatically Created Assets get switched on constantly, often years ago, often during a recommendations sweep that someone approved in three clicks. Campaign-level broad match is a different thing to keyword-level broad match, and plenty of teams have it on without realising it, because it was applied at the campaign settings layer, not the keyword layer.

The audit, before September

Do this in Google Ads Editor, not the web interface. It is faster and you can see the whole account at once.

1. 1.Filter every Search campaign for the campaign-level broad match setting. Note which ones have it on.
2. 2.Check asset-level settings for Automatically Created Assets. This lives with your responsive search ads, not in campaign settings, which is exactly why it gets missed.
3. 3.List every campaign that appears in either bucket. That is your September list.
4. 4.For each one, open the final URL and ask a plain question: if Google were free to send traffic to any page on this domain, which pages would embarrass us? Careers. Old promotions. Thin location pages. A support article that ranks well but sells nothing.

Final URL expansion is the component that carries real risk for a business with a large site. You can restrict it, and you should, before the upgrade happens rather than after. Nominate the URLs you are happy to send paid traffic to and exclude the rest.

Text customisation and brand control

Text customisation rewrites your ad copy per query. For a business with an approved messaging framework, legal review, or regulated claims, that is not a small thing. The control you have is upstream: the assets you supply and the landing page content it draws from. If your service page contains a claim you would not put in an ad, it can end up in an ad.

Two practical moves. Tighten the landing page copy for the pages in scope. And keep at least a few pinned headlines carrying the claim language you have signed off, so there is a floor under what gets generated.

Reporting, and the cadence Google itself recommends

AI Max reporting now segments search terms separately, search terms paired with the landing page AI Max chose, and the DSA equivalents, alongside keyword, asset and landing page reports. That is more visibility than the old DSA reporting gave you.

Google's own guidance tells advertisers to review search terms and item groups fortnightly, and to use negative keywords sparingly so you do not filter out matched traffic. That second instruction is a genuine change in operating posture and it deserves its own conversation, because it runs against fifteen years of habit.

How to sequence this

Do not wait for September and do not flip the whole account. Upgrade one campaign voluntarily, now. Pick a mid-volume campaign with clean tracking and a landing page set you are happy with, not your biggest revenue driver and not a dead one.

Run it for four to six weeks. Long enough to clear the learning period and cover at least one full purchase cycle. Then compare it to a similar campaign you left alone, controlling for seasonality. You are looking for three things: did volume rise, did efficiency hold, and did the query mix drift somewhere you do not want to be.

If it works, upgrade in waves and keep a control group untouched until the February 2027 date forces your hand. If it does not, you now have real evidence for your own account before the decision is taken away from you, which is worth more than any published percentage.

The honest limitation

You cannot cleanly A/B test this. Google's 7% figure comes from Google. Your before-and-after comparison is contaminated by seasonality, auction dynamics and whatever else changed that month. The best available read for most accounts is a matched-campaign comparison plus a geo holdout if your budget supports one.

That is a weaker measurement than anyone would like. It is still much better than finding out in September that fourteen campaigns changed behaviour overnight and nobody wrote down what the settings used to be.

Export your campaign settings to a spreadsheet this week and date it. That file is the only version of the truth you will have after the upgrade runs.