Most of the posts you will find about the DSA to AI Max migration are wrong, including several published by agencies who should check their sources.
They say September 2026. That was the original plan. In June, Google’s developer blog announced the automigration had been pushed to February 2027, and that the ability to create new Dynamic Search Ads campaigns, which had been switched off, was restored on 15 June 2026.
So if you read in April that your DSA campaigns were about to be force-upgraded and then nothing happened, that is why. You have roughly four months longer than you thought. The useful thing to do with them is not to relax, it is to run the test that the automatic upgrade will otherwise run on your live budget.
The actual timeline
| Date | What happens |
|---|---|
| 15 June 2026 | DSA campaign creation restored across UI, Editor and API |
| January 2027 | Ability to create new DSA campaigns removed |
| February 2027 | Automatic upgrade begins for remaining active DSA campaigns |
Sourced from the Google Ads developer blog, which is where timeline changes land first and where the marketing announcements tend not to get updated.
Worth noting that the original business announcement page still reads as though September is the deadline. If you are briefing a client from that page, you are briefing them on a schedule that no longer applies.
Find every affected campaign before you plan anything
If you have API or script access, this is the query. It returns every enabled campaign with a DSA configuration:
SELECT campaign.id, campaign.name, campaign.status, campaign.dynamic_search_ads_setting.domain_name FROM campaign WHERE campaign.status = ‘ENABLED’ AND campaign.dynamic_search_ads_setting.domain_name IS NOT NULL
Without API access, filter your campaigns list by campaign type and look for dynamic ad groups. It is slower and across a roster it is genuinely tedious, but it finds the same thing.
Do this first. A surprising number of accounts have one forgotten DSA campaign running at low spend that nobody has opened in a year, and those are exactly the ones that get auto-upgraded into something nobody is watching.
You have three destinations, not one
Google’s documentation names three paths for a DSA campaign: AI Max for Search, Performance Max, or a standard Search campaign using broad match with Smart Bidding.
The automatic upgrade picks for you. Choosing yourself is the entire value of the next four months.
Here is where I will disagree with most of the coverage: for a lot of local lead-generation accounts, the right answer is none of the three. DSA was often used as a crutch, a way to catch queries nobody had done the keyword research for. If that describes your account, the migration is a good reason to go and do that research properly and run a real Search campaign with keywords you chose, rather than swapping one black box for a slightly larger one. AI Max makes sense when your site is big, your catalogue changes, and query diversity genuinely outruns what you can maintain by hand. A plumber with eleven service pages does not have that problem.
Use the window to run the experiment
Google’s own recommendation, and the only one in the announcement I would repeat without qualification, is to set up Campaign Experiments and test AI Max against your existing DSA campaign rather than waiting.
The reason this matters is that February will give you a migrated campaign and no counterfactual. You will not know whether the new numbers are better, worse, or just different, because the thing you would compare against will no longer exist.
Before you start, export the last 90 days by campaign: spend, conversions, CPA, conversion rate, and the search terms that actually converted. Save it somewhere that is not the Google Ads interface. This is the single cheapest thing on this page and the one most likely to be skipped.
What AI Max switches on
Three features, and the names undersell what two of them do.
Search term matching finds queries beyond your keywords. Text customisation writes and rewrites your headlines and descriptions. Final URL expansion sends traffic to pages on your site that you did not nominate.
For campaigns migrating from DSA, all three come on by default, with legacy URL controls preserved. Dynamic ad groups become standard ad groups.
The controls are real but buried. Per Google’s setup documentation, search term matching sits under Settings, Campaigns, your campaign, then the AI Max section, and has to be on at campaign level before you can adjust it per ad group. Text customisation and final URL expansion live in the same section under Asset optimization. URL exclusions are at Campaign Settings, AI Max, Final URL expansion, then Add URL exclusions, with a Rules tab for pattern-based exclusions. Brand controls sit under Brands in the same section, and ad group brand inclusions override campaign level, which is easy to trip over.
(Brand lists themselves are managed somewhere else entirely, under Tools and the Shared library. No, I do not know why.)
The thing that makes this expensive
AI Max is a signal amplifier. You hand it a conversion action and it now has three more levers to pull in pursuit of it: which queries to chase, what the ad says, and where the click lands.
That is excellent when the signal is real.
A roofing account we took over in Australia reported 318 conversions at roughly a $10 cost per lead. 96% were taps on the phone-number button rather than connected calls. Real traceable leads came to 12 over sixteen months, and the true cost per acquisition was closer to $263. Smart bidding had spent over a year getting extremely good at finding people who tap phone buttons without calling. We pulled it apart in the 318-lead mirage.
Give that account text customisation and final URL expansion and nothing improves. The machine does not become more wrong, it becomes more efficient at being wrong, across more surfaces.
So before the experiment, open Goals, then Conversions, and confirm every primary action requires a real business outcome to fire. If it does not, fix that first, because everything downstream is optimisation toward a number that is not true.
Where the migration is riskier than average
Tight-margin lead generation. More reach is not automatically more profit. The trap is that the dashboard looks fine either way, because conversion count moves first and lead quality is the number nobody measures for another six weeks. Ask whoever answers the phone what changed in the enquiries before you ask the account. If they say the calls got vaguer and further away, you have your answer regardless of what the CPA column says. The same credit-attribution problem we covered in who really earned the lead applies here.
Regulated verticals. Text customisation rewriting your headlines is a compliance question in legal, medical and financial accounts, not only a performance one. Somebody has to read what it wrote, every week.
Large or untidy websites. Final URL expansion on a small brochure site is low risk. On a site with thousands of URLs, an old campaign directory and an unmaintained blog, it is a spend leak until you exclude properly.
Where I am still not settled
Whether February 2027 is the final date.
This deadline has already moved once, from September 2026, after advertiser pushback. Planning a client roster around a date Google has already revised is uncomfortable, and I would not be shocked to see it move again. That argues for migrating on your own schedule rather than building a plan that depends on the deadline holding. It also argues against telling clients the sky falls in February, because if it slips again you have spent credibility you did not need to spend.
What to actually do this quarter
Find the DSA campaigns. Save a 90-day baseline before there is nothing to compare against. Fix the conversion actions. Then run one Campaign Experiment against your largest DSA campaign and let it settle for a full cycle, so that when February arrives you are confirming a decision rather than discovering one.
If you are not sure which of your campaigns are affected, or the numbers have already moved and nobody can tell you why, that is the kind of thing we take apart in an audit. Get a free audit and we will map what is migrating and what it is likely to cost you, or read why cost per lead climbs while the dashboard still looks healthy.
AI Max is not the only forced migration on the calendar. If you run pay-per-lead campaigns, the Local Services Ads move into Performance Max follows the same pattern, including the reporting you lose if you do not export it first.
Researched and drafted with AI assistance, then reviewed, fact-checked and edited by Rajat Verma before publishing. All account examples are from real campaigns managed by Turquoise Digital and are anonymised.