From DSA to AI Max: the complete migration guide for September 2026

DSA stands for Dynamic Search Ads: a campaign type in Google Ads where Google decides the search terms itself and generates ad headlines from the content of your website. You select no keywords by hand, the system optimises automatically. Which is what made DSA popular for years. Many agencies use DSA alongside regular campaigns, and some run their whole strategy on it.

The core problem: many business owners do not even know their campaigns run on DSA. Your monthly report shows results, but not the technical foundation under the campaign. And it is exactly that foundation which is disappearing.

The date is fixed, and the ending is milder than it sounds. You read everywhere that Google switches off all DSA campaigns in September 2026, with no wind down and no warning. That is not right. Google writes that the phase out and the automatic conversion begin in February 2027, that in September 2026 only campaigns with automatically created assets and cross ad group broad match go over, and that ad groups with dynamic targets are converted into standard ad groups with settings and data carried over to AI Max. Source: Google on the upgrade from DSA to AI Max. Last checked on 25 August 2026: the post is still there and February 2027 has not moved again.

Google is replacing DSA with AI Max, a system built on large language models, comparable to the technology behind ChatGPT. It reads your website live, understands the context of what you sell and decides in real time which search is commercially worth bidding on. That approach differs fundamentally from DSA, which mainly matched text from your pages to search queries.

The main differences, side by side:

• Real time optimisation on the chance of a conversion rather than static matching
• Dynamic ad copy adjusted per search query
• Better audience selection based on context and behaviour

There is a commercial side to it as well. AI Max earns Google more: smarter targeting means higher click prices on more relevant searches. The step fits a broader movement in which Google pushes AI into every part of search and advertising. To see how far that already goes, read how you get found in Google AI Overviews and ChatGPT, or how Google AI rings on behalf of your customer to book an appointment.

You do not need to be a Google Ads expert to check this. Send one email to your marketing agency with the question: is my Google Ads campaign running on DSA, and if so, what is your plan for the AI Max migration?

Then pay attention above all to the kind of answer you get back:

• Good answer: a concrete migration plan with dates, for example testing in the autumn and switching over fully in December.
• Poor answer: "It will be fine, we are keeping an eye on it" or a vague technical explanation with no plan.

Do you manage your campaigns yourself and are you unsure? We run a free DSA check for you. Within our service AI and automation we follow this kind of shift at Google daily.

The move to AI Max touches three parts of your campaign directly.

Click prices

What a click is worth is decided in real time. That means you will pay more for some searches, but you will also win a customer more often. Expect the first weeks after migration to be unstable. Anyone who promises you a smooth switch has not run one.

Search queries

The targeting in AI Max is finer than in DSA. The system looks at search behaviour, context and the chance of a conversion. That means your campaign can appear on searches you did not expect. So check your search terms report more often than you are used to.

Conversion patterns

The route from click to customer changes as well. If your conversion tracking is not set up properly, AI Max will optimise towards the wrong things. So put that right before you migrate a single campaign.

Plan the migration in four steps and work towards February 2027, the month in which Google converts the rest itself:

• Now, take stock and make a plan: check which campaigns run on DSA and ask your agency for a written migration plan.
• Autumn 2026, a test campaign: start one AI Max test campaign alongside your existing DSA and compare the results for two weeks.
• December 2026, the full switch: migrate every campaign on the basis of your test results and give each campaign at least a week to settle. From January 2027 you can no longer create a new DSA.
• February 2027, the automatic conversion: anyone who migrated in December did the conversion themselves instead of having it done to them.

Warning: wait until January 2027 to start and you miss the test phase entirely, because you can no longer create a new DSA to test alongside.

Where the dates come from, and where they do not. February 2027 is in Google's own post on the upgrade from DSA to AI Max: the phase out and the automatic conversion begin in that month. January 2027, the month in which you can no longer create a new DSA, is not in that post; it comes from the explanation by Google Ads Liaison alongside the same delay announcement, recorded by Search Engine Land. Last checked on 25 August 2026: both are still there and neither date has moved again. The rest of this planning is ours, not Google's.

Migrating yourself is possible, but be honest about what it takes. At the very least you need:

• A working Google Ads account with conversion tracking set up correctly
• Access to Google Analytics
• Two to three weeks of time for daily monitoring

On top of that, count on at least four weeks of learning time before the new system understands your market. A mistake in the migration soon costs you weeks of revenue, and that is exactly why many companies hand it over. Would you rather read on yourself about how Google Ads and AI are developing, then there are more articles in our knowledge base, which is in Dutch.

Two English pieces sit closest to this one, because they cover the other two things Google changed in the same year: the target CPA change of 17 August 2026, which touches every campaign that is limited by budget, and the budget pacing change of June 2026, which moves what a daily budget adds up to in a month. Decide to hand the migration over, and what that looks like in practice is described under outsourcing Google Ads to an agency.

These four mistakes are the ones we see most often in migrations to a new campaign type:

• Starting too late. The most common mistake: agencies are overloaded in December and you are at the back of the queue.
• Moving everything at once. Test one campaign, learn from it, and only then move the next.
• Not checking conversion tracking. This is fundamental. Wrong tracking leads to optimisation on the wrong signals.
• Trusting your agency blindly. Ask for a migration plan and for interim results, even when the relationship feels good.

You can only judge a migration if you know where you came from. That sounds obvious, and yet this is the part that gets skipped most often: the campaign is converted, things run differently for a few weeks, and then the argument starts about whether it is better or worse without anyone still holding a sharp picture of the old situation. So record a baseline before the conversion, and record it outside Google Ads, in a spreadsheet or a document that belongs to you.

What goes into that baseline is not a matter of taste. Take the same period for every campaign that runs on DSA, preferably the last thirteen weeks, and write down what you will want to compare later:

• The four basic numbers per campaign. Impressions, clicks, cost and conversions, plus the cost per conversion that follows from them. Note the period you used, otherwise in six months you will be comparing apples with a different month.
• The search terms report, exported. This is the most valuable file of the whole migration. It shows which real searches your DSA came out on and what they produced. After the conversion it is your only way to see whether AI Max hunts on the same ground or on new ground.
• Your exclusions. The negative keyword lists, the excluded pages or audiences, and the competitor brand names you deliberately filter out. These lists are often built up over years and nobody remembers why everything is in there. They are still needed after the conversion, and sometimes needed harder than before.
• The ad copy that worked. Copy your best performing headlines and descriptions into your document. The system may adjust them from now on, but you want to be able to look back at the tone that brought your customers in.
• The seasonal pattern. If your market peaks in December and goes quiet in July, that belongs in your baseline, otherwise you will read a season as a migration effect later on. This is the most common thinking error after any switch: the calendar gets blamed for the system, or the system gets blamed for the calendar.

This baseline costs an hour. It is the difference between being able to say in three months what happened, and having a feeling in three months. It is also your only brake on a conversation that would otherwise go in every direction: whoever puts figures from before the conversion on the table is arguing about facts instead of impressions.

Of everything you can do before the migration, there is exactly one thing you must not skip: checking whether your conversion tracking is correct. The reason is simple. AI Max steers on the chance of a conversion. A system that optimises on conversions is only as good as the signal it gets. Measure the wrong things and the new system will optimise faster and harder towards the wrong side than the old one did, and you will only notice when your invoice and your diary stop agreeing with each other.

Go through these five points in your Google Ads account, or have them checked and ask for a written answer per point:

• Which actions are set as a primary conversion? Google Ads separates primary and secondary conversion actions. Only the primary ones steer the bidding. If a page view, a click on your phone number and a submitted form are all three switched on as primary, the system counts a visitor who did nothing just as heavily as someone who left their details.
• Is anything counted twice? An enquiry registered both through the form and through a thank you page counts twice. That looks harmless until the moment the system starts deciding which searches it considers worth more.
• Is the value you pass on correct? If all your enquiries get the same value, no system can tell the difference between a quote request of five hundred pounds and one of fifty thousand. For companies with widely varying job sizes, passing a realistic value per type of enquiry is the biggest improvement you can make, bigger than the migration itself.
• Is telephony included? At many local businesses the majority of enquiries come in by phone. If that is not measured, the system is blind to your most important channel and optimises purely on the smaller stream of form fills.
• What does your cookie banner do to the measurement? Visitors who refuse are not measured, or measured differently. That is not a fault but a given, and it belongs in your expectation: it explains part of the gap between what Ads measures and what lands in your inbox.

Find a fault here, and you repair it before the conversion, not after. A bidding strategy that has learned through a period with broken measurement drags those weeks along with it. Google Ads has a setting to exclude a period from what the bidding strategy learns, and that setting exists for exactly this kind of accident. Better still is not to need it. How the measurement side and the findability side touch each other is worked out under why you score low on PageSpeed, where the same principle applies: a number is only useful once you know what it is measuring.

The day of the conversion is not the end of the migration, it is the start of the only period in which you can genuinely steer. In those first two weeks the system works out which searches belong to what you sell, and the choices it makes then carry on working for months. A fixed rhythm helps, and it costs less time than people think.

Day one to three: only look, do not turn anything

Look at your spend and your search terms daily, and change nothing about your budget or your bidding strategy. The temptation to intervene on day two is strong, because the figures look odd. It comes with the territory. What you do do in these days is exclude what obviously does not belong to you: searches from job applicants, from people looking for a manual, from students writing an essay. Filtering that kind of traffic out is not steering, it is tidying up. Nobody ever lost a campaign by excluding job applicants.

Day four to seven: the search terms next to your baseline

Put the new search terms report next to the report you kept before the conversion. Three questions: are the searches that used to bring you customers still coming through, which category is new, and does that new category produce anything or only cost? New searches are not automatically bad. They are the whole point of the new system. They should, however, produce something within a few weeks.

Day eight to fourteen: the first real judgement

Only now may you have an opinion on the cost per enquiry, and even then with a pinch of salt. Do not compare with the best week you ever had, but with the average from your baseline over the same kind of period. If the counter stands less favourably, look first at whether it is the quality of the enquiries or the number. More enquiries worth less asks for a different repair than fewer enquiries of the same kind.

And agree in advance when you will roll back. A limit you write down beforehand, for example that you intervene if the cost per enquiry is still well above your baseline after three weeks, is a decision. The same limit thought up halfway through is a reaction to a bad week. Those two feel the same and work out very differently.

The advice to test before you convert applies to everyone. What you watch does differ per type of business, and that is rarely split out in the general stories about this migration.

• Webshops with many products. Here DSA was often used to cover the long tail of product pages you would never type in by hand. When converting, watch above all your sold out and discontinued products: a system that reads your site live can send traffic to pages that no longer produce anything. Make sure your page structure and your stock status are correct before you convert.
• Service businesses with a small offering. Five services, five pages, and DSA as a safety net next to your keyword campaigns. For you the question is mainly whether the new system will start competing with your own existing campaigns. Keep an eye on whether your brand name and your core keywords are suddenly being served from two campaigns at once.
• Local trades with a working area. For you the geographic boundary matters more than the keywords. After the conversion, check whether your adverts stay inside your driving area and whether enquiries from outside that area are quietly increasing. How to handle that local side is worked out under Performance Max for trades, which deals with the same tension between an automated system and a fixed working area.
• Businesses with long sales cycles. If weeks pass between the first click and the order, two weeks of measurement is too short to conclude anything. Take the number and the quality of the first conversations as an interim measure here, and agree that the real judgement only falls after a full sales cycle.

In all four cases the same underlying rule holds: the clearer your website says what you do, for whom and where, the less the new system has to guess. A system that reads your pages live is, after all, only as sharp as the pages it reads. That makes the migration a good moment to go through your most important service pages again, because from now on that text is doing double duty.

What happens if I do nothing?

Your adverts do not disappear. Google converts your dynamic ad groups into standard ad groups and carries the settings and the data over to AI Max. The difference is that Google then makes the default choices, at the moment that suits Google, and that you are correcting course in the first weeks after the conversion instead of testing beforehand.

Is AI Max more expensive than DSA?

Not by definition, but click prices are sometimes higher. The first weeks after migration are usually dearer than the period after that.

Can I leave my DSA campaign running until the conversion?

You can, but then you miss the test period. It is better to start a test campaign alongside your existing DSA in the autumn of 2026, because from January 2027 you can no longer create a new DSA to test against.

How long does the migration take?

Technically the switch costs a few hours. Optimisation and settling down cost two to four weeks per campaign.

Do I have to change my website?

Not necessarily, but the clearer your content, the better AI Max can target. After all, the system reads along on your site live.

You do not have to carry out this article in an afternoon. The order matters more than the pace: first find out whether your campaigns run on DSA, then go through your conversion tracking, then record your baseline, and only after that convert. Anyone who keeps that order can sit back in February 2027 and watch Google convert the rest, because by then there is nothing left to convert. Anyone who reverses it migrates in the dark.

If you are not sure whether your account is one of the affected ones, or you want someone to look at the search terms report with you before you switch anything on, put that question to us. Name your sector, the number of campaigns and whether you manage them yourself or through an agency, and we will look at your set up and say honestly when you are better off doing it yourself. That goes through the contact page, where you can pick a moment in the diary straight away. If you would rather hand the account over completely, our approach is described under outsourcing Google Ads, and what it costs is on the pricing page.