Last Updated on September 3, 2026
Between 1 and 30 September, Google is upgrading Search campaigns to AI Max automatically. You do not opt in. If your account uses campaign-level broad match or standalone automatically created assets, you are on the list.
Most of the advice you will read this month is about which box to untick. That advice is aimed at the wrong problem.
What is actually happening
Google has set a timeline for retiring older Search campaign features and folding them into AI Max, its automated targeting and creative suite. The dates below were reported by Search Engine Land from Google’s Ads Developer Blog announcement.
| Date | What changes |
|---|---|
| 3 August 2026 | Google stopped allowing new campaign-level broad match and legacy automatically created assets to be created, across the interface, Ads Editor and the API. |
| 1 to 30 September 2026 | Automatic migration of those campaigns to AI Max, rolled out gradually through the month. |
| September 2026 | In-account notices start encouraging advertisers to move Dynamic Search Ads over voluntarily. |
| February 2027 | Dynamic Search Ads migrate automatically, and new DSA ad groups can no longer be created. |
Google says campaigns will be migrated in place using equivalent AI Max settings, and that existing brand inclusions and exclusions carry over automatically.
The number Google leads with, and what it actually compares
Google’s public claim is that advertisers who activate AI Max “will typically see 14% more conversions or conversion value at a similar CPA/ROAS”, rising to 27% for campaigns still mostly using exact and phrase keywords.
Now read the footnote. Google states the figure comes from its own internal data from 2025, based on campaigns with more than 70% of conversions or conversion value coming from exact or phrase match keywords, and for non-retail advertisers only.
So the 27% is not a forecast for your account. It is the gap measured on accounts that were doing almost no automated matching at all, with retail taken out. If you already run broad match, Google’s own footnote is telling you that you are not the business in that number.
The Whito rule on vendor statistics: before you use a number, find out what it compares. An uplift measured against a business doing none of the thing is not a forecast for a business already doing some of it.
What happened when somebody else counted
Smarter Ecommerce, a Google Ads tooling firm, analysed more than 250 Search campaigns running AI Max and published the results in March 2026. The median campaign produced 13% more conversion value, close to Google’s non-retail claim.
The median campaign also produced a cost per acquisition 16% higher.
On return on ad spend the median difference was zero, but the spread ran from 42% above the campaign’s own baseline to 35% below it. Their description was a coin toss.
That is not a scandal. It is how paid advertising works. The next conversion costs more than the last one. It is only a problem if nobody budgeted for the next conversion costing more, which is exactly what happens when a business has never worked out its numbers. Ours are set out in our breakdown of where the UK’s £46.7 billion advertising spend actually went.
Why this lands harder on small accounts
Automated bidding learns from conversion data. A national retailer feeds it thousands of conversions a month. A plumber in Leeds feeds it eleven. The wider the range of possible outcomes, the more that thin data matters, and small accounts sit at the wide end of the range.
There is a second problem, and it is the one that costs real money. The system optimises towards whatever you told it a conversion is. If your only conversion is a click on the phone number, you will buy phone clicks. Suppliers, wrong numbers and people wanting a job you do not do all count as wins. Google gets more efficient at buying you the wrong thing.
What to do before the end of September
Three things, in order. Do not skip to the third.
1. Work out what a customer is actually worth. Not a lead, a customer. Job value, minus costs, times your conversion rate from enquiry to sale. Until you have that figure, no bidding setting can be judged good or bad, because you have no threshold to judge it against. We have done the arithmetic for twelve common jobs in our break-even lead prices for UK trades.
2. Fix the signal before Google changes the engine. Make your conversion the thing that makes money, not the thing that is easy to count. Put a negative keyword list in place. Set your brand exclusions so the migration carries something useful across. Check where your budget sits relative to your turnover, because the sensible range for a sole trader is not the range an agency will quote you.
3. Then let it run, and compare it to your own baseline. Not to Google’s 14%, and not to a case study about L’Oreal. Your cost per customer last quarter against your cost per customer this quarter. That is the only comparison that pays your wages.
The takeaway
Google is not removing your control this month. It is removing the part of your control that was doing the least work. A keyword list was never the thing protecting your budget.
An account with honest conversion tracking and a real cost per customer will come through this migration and probably learn something. An account without one was already losing money quietly. From September it will lose it faster, and with better reporting.
