Dynamic Search Ads end in February 2027, not September. Here's the timeline that applies now

Google has moved the Dynamic Search Ads migration twice, and most of the advice currently ranking for this topic is working from the superseded date. If your plan says September 2026, it is out of date by five months — which is either a problem or a gift, depending on what you do with the window.

The timeline that actually applies, because it changed

If you search this topic you will find a lot of pages telling you that Dynamic Search Ads end in September 2026. They were right when they were written. Google pushed the migration back after advertiser feedback, and the dates that matter now are January and February 2027.

That is the single most valuable thing on this page, so here it is as a calendar rather than prose.

One clarification worth making early, because the two changes get conflated constantly. The September 2026 migration of campaign-level broad match and automatically created assets is not this. That one already ran, and we covered what to audit in the AI Max auto-upgrade pre-migration checklist. This page is about DSA campaigns specifically, which got their own extension and their own February 2027 date.

What you are actually being moved to

AI Max for Search is out of beta. Google reports that campaigns using the full feature suite see, on average, 7% more conversions or conversion value at a similar CPA or ROAS. Treat that the way you would treat any platform-reported average: it is a real signal that the product works for a lot of accounts, and it is not a forecast for yours.

The structural change is more informative than the headline number. DSA solved one problem: filling the gap between your keyword list and the pages on your site, by matching queries against page content. AI Max folds that capability into the ordinary Search campaign and adds query expansion beyond your keywords, asset generation from the landing page, and a set of controls DSA never had — brand inclusions and exclusions chief among them.

So the migration is not like-for-like in either direction. You gain controls you have been asking for. You also lose the clean mental model where a DSA campaign was a visibly separate bucket you could budget, read and switch off on its own.

The lever-by-lever audit

This is the work. Before you migrate anything — voluntarily or by waiting for February — go through every DSA lever you currently rely on and establish what carries it afterwards. Do not assume the mapping; open the campaign and check.

Migrate now or wait for February? Score it

There is no universally right answer, so stop looking for one. Score your account out of 10. One point for each statement that is true.

1. DSA accounts for more than 20% of the campaign's conversions.
2. Your site has more than about 200 indexable pages.
3. You have pages you specifically do not want to receive paid traffic.
4. Brand terms are a meaningful share of the account's volume.
5. You run in more than one language or country from the same campaign.
6. Your landing page copy is out of date on more than a handful of templates.
7. You have offline or delayed conversions feeding the account.
8. Q4 is your peak trading season.
9. Nobody will be watching the account closely in February.
10. You have never configured brand inclusions or exclusions.

Note what points 8 and 9 do to the timing. If Q4 is your peak, you do not want to be learning a new campaign type in November, and you also do not want an automatic upgrade landing in February while the person who understands the account is away. The window that satisfies both is January, which is also the month new DSA creation disappears. That is not a coincidence — it is the last month where you can still stand up a DSA campaign as a control if the test goes badly.

How to test it properly, with the arithmetic

If you scored 7 or higher, run a real comparison rather than switching and squinting at the numbers afterwards. The design is simple; the part people get wrong is the duration.

1. Split by campaign, not by ad group. Duplicate the DSA campaign, upgrade the copy to AI Max, and run both against the same budget split.
2. Hold the variables still. Same bidding strategy, same target, same conversion action, same geo. If you change the bid strategy at the same time you will never know which change did what.
3. Let the learning period finish before you read anything. Roughly two weeks, longer if your conversion cycle is long.
4. Size the window from your conversion rate, not from a calendar. Aim for at least 50 conversions per arm before you compare.

The arithmetic that decides whether the test is even possible: if the DSA campaign currently produces 3 conversions a day and you split 50/50, each arm gets about 1.5 a day. Fifty conversions per arm is 34 days. Add the two-week learning period and you are planning a seven-week test. Start that in mid-October and you are reading results in early December, which is both too late to act calmly and contaminated by peak-season traffic.

Which gives you the real decision rule. If your campaign cannot generate 50 conversions per arm inside seven weeks, do not run the test. You will not get a readable result, and the honest move is to migrate in January with a careful configuration and watch it weekly instead. Under-powered tests are how accounts end up making a permanent decision on a fortnight of noise.

The five things to do before January

1. Export everything. Current DSA settings, dynamic ad targets, page feed, negative lists and 12 months of performance. After an in-place upgrade, the old configuration is not there to consult.
2. Set brand inclusions and exclusions now. They carry over on migration, which means whatever is there in February is what you get. Empty carries over as empty.
3. Fix the landing pages the generator will read. Text customisation writes from page copy. Stale headlines and missing H1s become stale ads.
4. Prune your URL set. Decide explicitly which pages may receive paid traffic and enforce it with the page feed and URL exclusions rather than hoping.
5. Put a named owner and a date in the calendar for February 2027. Automatic migrations are not the problem. Unattended automatic migrations are.

Frequently asked questions

When exactly do Dynamic Search Ads stop working?

Existing DSA campaigns keep running until February 2027, when Google automatically upgrades the remaining ones in place to AI Max for Search using equivalent settings. From January 2027 you can no longer create new DSA campaigns in the Google Ads interface, Google Ads Editor or the Google Ads API. In-account notices prompting a voluntary upgrade began in September 2026.

Was the DSA to AI Max migration delayed?

Yes. The automatic upgrade of DSA campaigns was originally scheduled for September 2026 and Google pushed it back to February 2027 following advertiser feedback asking for more time. The ability to create new DSA campaigns, which had begun to be phased out, was also restored in mid-June 2026 and now ends in January 2027.

Is this the same as the September 2026 AI Max upgrade?

No, and conflating them causes real mistakes. Campaign-level broad match and automatically created assets migrated to AI Max during September 2026 on their own schedule. Dynamic Search Ads are a separate campaign type with a separate, later deadline of February 2027. If your account was affected by the September change, that has already happened and is unrelated to your DSA timeline.

What replaces the DSA page feed in AI Max?

The page feed becomes the URL source for final URL expansion. The critical check after migration is whether expansion is drawing only on your feed or has become eligible for pages you never approved. That is where discontinued products, thin category stubs and old landing pages start receiving paid traffic, so pair the feed with explicit URL exclusions rather than assuming the feed still acts as a hard boundary.

Should I upgrade to AI Max now or wait for the automatic migration?

Score the account. If DSA drives under 20% of conversions, your site is small and you have no pages to protect, waiting for February is fine. If DSA is a major conversion source, your site is large, brand terms matter or Q4 is your peak, migrate deliberately in January after peak trading, with brand inclusions and exclusions configured first and the page set pruned.

How long should an AI Max versus DSA test run?

Long enough to reach roughly 50 conversions per arm after the learning period ends, which takes about two weeks. At 3 conversions a day split evenly, each arm earns 1.5 a day, so 50 conversions takes 34 days and the full test is about seven weeks. If your campaign cannot reach that inside seven weeks, skip the test and migrate carefully in January instead, because an under-powered test produces a confident answer from noise.

The takeaway

The delay is the story, and it is worth more than the migration advice attached to it. You have been handed five extra months on a change that was going to happen to most accounts whether or not anyone was watching. Use them in this order: export the current configuration while it still exists, set the brand controls that will carry over, decide explicitly which URLs are allowed to receive paid traffic, and fix the landing page copy that the asset generator is about to read aloud. Then pick your own date in January rather than inheriting Google's in February. The accounts that come through platform migrations well are almost never the ones that found a clever setting. They are the ones that arrived with the configuration already decided.