DSA to AI Max Migration: The 2026 Playbook for Search Campaigns
Dynamic Search Ads are being retired. From September 2026 Google starts auto-upgrading DSA campaigns to AI Max, new DSA campaigns can no longer be created, and the final sunset has been pushed out to February 2027. This is not a rename. The way queries are matched changes, the way landing pages are chosen changes, and the controls you spent years building do not all survive the move untouched. This guide covers what actually breaks during an automatic upgrade, how to run the DSA to AI Max migration on your own terms, what to snapshot before you touch anything, and which numbers tell you within two weeks whether the move helped or hurt.
The timeline, and why it matters
DSA shipped in 2011 on a simple promise: point the system at your site, let it crawl your pages, and it will match user queries to the right URL and write the headline itself. Your controls were dynamic ad targets — page categories, URL rules, page feeds — plus negative keywords.
AI Max works differently. It is not a campaign type. It is a set of features you switch on inside a standard Search campaign, at campaign or ad group level. There are three of them: search term matching, text customization, and final URL expansion. That last one is the direct descendant of DSA.
| Window | What happens |
|---|---|
| Through August 2026 | Voluntary upgrade via the in-product tool. You keep full control of how settings transfer |
| September 2026 | No new DSA campaigns can be created. Automatic upgrades begin |
| End of September 2026 | All eligible campaigns are expected to have been upgraded |
| February 2027 | Final sunset. Originally set for September 2026 and extended after advertiser feedback |
The extension is not permission to wait. Campaigns you do not upgrade yourself will still be upgraded — with Google’s defaults, which means all three AI Max features on. The extra months are for a controlled migration with measurement, not for procrastination.
What actually changes when you move
The difference is the signal source. DSA leaned almost entirely on your site content: whatever the crawler found on a page determined which queries it could match. AI Max adds real-time intent signals — how the query is phrased, what the session looks like, what the user was doing before. Three practical consequences follow.
Reach widens, and not always in your favour
DSA brought traffic that looked like your pages. AI Max brings traffic that looks like intent to buy what you sell. The overlap is large, but at the edges you will see queries DSA never touched: broader, higher-funnel, sometimes competitive. Some of them convert better than anything you had. Some are pure budget leakage until you cut them.
Your ad copy stops being entirely yours
Text customization rewrites headlines and descriptions against the specific query, drawing on your assets and landing page content. For a retailer with thousands of SKUs that is a clear win. For regulated categories, where the exact wording of a claim carries legal weight, it is a risk: the system can assemble a headline you would never have signed off. The feature can be disabled independently of the other two.
URL expansion needs explicit rules
Final URL expansion inherits DSA’s job. If you spent years building exclusions so the crawler stopped sending paid traffic to your careers page, your privacy policy, or out-of-stock product pages, those exclusions have to survive the move. The upgrade tool carries URL controls across, but not always completely. Verifying them is the single most important post-migration check.
Snapshot everything before you migrate
The expensive mistake is upgrading and then having nothing to compare against. Pull a 60-day baseline at campaign and ad group level before you touch a single setting:
- Cost, clicks, conversions, cost per conversion, conversion value, ROAS.
- Top 200 search terms with cost and conversions, exported from the search terms report and negative keyword workflow. This is how you will later identify exactly what AI Max added.
- Every dynamic ad target: which categories, which URL rules, which exclusions.
- Every negative keyword list, at campaign, list and account level.
- Spend distribution by landing page, so you can see immediately if budget starts flowing to different URLs.
- Impression share and lost IS to budget and rank — your reach baseline. How to read those figures is covered in the piece on impression share and auction insights.
Also record which bid strategy each campaign runs and at what target. There is a second change running in parallel: since 17 August 2026, budget-limited campaigns hold their CPA and ROAS targets far more tightly instead of overdelivering against them. Migrate in the same window and you risk blaming the migration for something bidding did. The separation is explained in the breakdown of the August 2026 target-based bid strategy change.
Running the DSA to AI Max migration manually
Step 1: sort campaigns into three buckets
Do not move everything at once. Sort by risk:
- Low risk. Modest spend, broad catalogue, no compliance sensitivity around wording. Move these first and learn the process on them.
- Medium risk. Meaningful spend, stable CPA, accumulated URL exclusions. Move second, always with an experiment running.
- High risk. The bulk of account spend, regulated vertical, strict copy requirements. Move last, and only after the first two buckets have run clean for three to four weeks.
Step 2: use the upgrade tool, do not rebuild
A banner and a dedicated upgrade tool appear in the interface. It ports historical settings and data into standard ad groups and preserves accumulated learning. Rebuilding the campaign from scratch throws that away and hands you an avoidable ramp-up period. It is the worst available option.
Step 3: work the post-upgrade checklist immediately
- URL controls. Open final URL expansion settings. Confirm every DSA exclusion made it across. Manually re-add anything missing: utility pages, blog (if it should not take paid traffic), cart, account area, out-of-stock pages.
- The three features. All are on by default. Decide on each separately. A conservative starting point: search term matching on, URL expansion on with hard exclusions, text customization off until you test it deliberately.
- Negatives. Move every list across. Pay particular attention to account-level negatives — they work, and they are the ones people forget to verify.
- Locations and languages. Confirm geo targeting did not reset to presence-or-interest. This is the classic leak after any bulk campaign edit.
- Assets. AI Max leans on your headlines and descriptions harder than DSA did. Weak assets produce weak customization. Aim for 12–15 headlines and three to four descriptions per RSA, and pin anything that must not be altered. Testing discipline is covered in the guide to RSA ad testing.
Step 4: freeze bids and budgets for 14 days
The system is rebuilding its matching model. Any target edit in that window mixes two effects and leaves you unable to attribute either. If a budget change is unavoidable, keep it under 20% and log the date.
Step 5: read search terms every two or three days
This is the one task you cannot defer. Open the search terms report, sort by cost, look for themes that were not in your baseline export. Cut irrelevant ones with negatives immediately, before they consume a week of budget. After two weeks, a weekly cadence is enough.
Measuring whether it worked
A before-and-after comparison of two-week windows almost always lies: seasonality, competitive shifts and parallel account changes generate more noise than the migration generates signal. Run an experiment instead.
If volume allows, split traffic 50/50: control on your current configuration, treatment on AI Max with your chosen settings. Setup and interpretation are covered in the guide to Google Ads experiments. As a rough guide, run for at least three to four weeks or until each arm has accumulated 50–100 conversions; below that, any difference you see is indistinguishable from noise.
If you cannot run an experiment, watch a set of metrics rather than one:
| Metric | What an increase after migration means |
|---|---|
| Impressions and clicks | Reach expanded. On its own, tells you nothing |
| Share of spend on new search terms | The key migration metric. Rough guide: 20–40% new is normal, over 60% needs investigation |
| CPA and ROAS on new terms only | This is where the actual answer lives |
| Conversions on landing pages that never got traffic before | Either URL expansion found working pages, or it is sending traffic somewhere useless |
| Impression share lost to budget | A sharp rise means AI Max found more demand than you are funding |
Rather than rebuilding this view by hand every week, put the segments you care about into calculated columns inside the interface — the method is covered in the guide to custom columns and the Report Editor.
Five ways this goes wrong
Waiting for the automatic upgrade
Auto-upgrade applies defaults: all three features on, URL controls ported as-is. If your DSA setup had any non-trivial rules, you will learn about the problem from the invoice rather than from a notification.
Leaving text customization on in a regulated vertical
Finance, healthcare, education, legal services — anywhere a claim carries weight. A generated headline can promise something you did not promise. Start with the feature off, test it separately, and audit what actually served.
Failing to move URL exclusions
The most expensive mechanical error. Budget drifts to blog posts, careers pages and help articles, conversions do not follow, and the interface looks fine because spend is still happening. The landing page report exposes it in about a minute.
Judging on three days of data
The first three to five days after migration are exploration: the system is probing new query space, and CPA is almost always worse. Rolling back on that sample guarantees you never escape a cycle of permanent rebuilds.
Migrating at peak season
If you have real seasonality, do not move your main spend during your highest-demand week. The model’s ramp-up will coincide with the moment mistakes cost the most.
Large accounts
With dozens of DSA campaigns, clicking through settings one by one is not realistic. Work in bulk: export the structure, build a sheet of campaign, spend, URL exclusions, negatives and bid strategy, mark risk buckets in the sheet, then apply changes en masse. The mechanics are in the guide to bulk operations in Google Ads Editor.
Separately, verify your conversion signal is intact. AI Max depends on signal quality more than DSA did: if enhanced conversions are misconfigured, or you feed the system a single generic goal with no value differentiation, expanded matching will optimise toward a poor target and buy a lot of cheap, useless traffic.
Who wins and who struggles
The pattern across accounts is fairly consistent. Broad catalogues win: retail, marketplaces, services with many landing pages covering distinct intents. URL expansion and intent matching genuinely surface demand your keyword list never described.
Narrow product sites of five to ten pages tend to lose. There is nothing to expand into, and the system starts sourcing adjacent traffic with no real relationship to the product. For those accounts the working approach is not to treat AI Max as the engine, but to keep the core on exact and phrase keywords and use AI Max as a controlled incremental source with aggressive negatives. How that core should be split is covered in the piece on keyword match types.
A third category is accounts already operating close to their trust and policy limits. Any reach expansion means more surface area for review, so the migration deserves a slower pace and more conservative settings. If you operate through Google Ads agency accounts, agree in advance which AI Max features you are enabling and on which campaigns — it removes a class of avoidable friction at review time.
Migration checklist
- Export a 60-day baseline of metrics and search terms.
- Save every dynamic ad target, URL rule and negative list to a separate file.
- Sort campaigns into three risk buckets.
- Upgrade the first bucket with the native tool. Do not rebuild.
- Work the five-point settings checklist: URL controls, the three features, negatives, geo, assets.
- Freeze bids and budgets for 14 days.
- Read search terms every two to three days for the first two weeks.
- Run an experiment on the second bucket for three to four weeks.
- Move your main spend only after a confirmed result.
- Record the new baseline and return to a normal optimisation cadence.
If you want a second pair of eyes on a specific account before migrating — which campaigns to move first, what to disable, where budget leakage is most likely — that is the kind of review where an hour saves a month of watching charts. Foundations are covered in the Google Ads guide, and account-level work is covered by the Google PPC agency service.
FAQ
When exactly do DSA campaigns stop working?
Automatic upgrades begin in September 2026, which is also when new DSA campaign creation ends. The final sunset is set for February 2027, extended from the original September 2026 date after advertiser feedback.
Can I opt out of AI Max entirely?
You cannot opt out of the DSA sunset. You can control what the campaign looks like afterwards: the three AI Max features toggle independently, and disabling URL expansion leaves you with a standard keyword-driven Search campaign.
Does campaign history and learning survive the upgrade?
Yes, if you use the native upgrade tool, which ports settings and data into standard ad groups. Manually rebuilding the campaign does not preserve history and restarts the learning period.
What happens to my DSA URL exclusions?
URL controls are carried over during the upgrade, but verify them manually. This should be the first thing you open post-migration: if exclusions did not transfer, budget starts flowing to utility pages immediately.
Should I adjust bids during the migration?
No — ideally change nothing for 14 days, or you will not be able to separate migration effects from your own edits. Also account for the 17 August 2026 bidding change if it overlaps your migration window.
Is text customization worth enabling?
For broad catalogues, usually yes — relevance improves. For regulated verticals or anywhere claim wording carries legal weight, start with it off, test it deliberately, and audit which copy actually served.
How much will reach grow after migration?
It varies with how broad your site is. As a rough guide, the share of spend going to search terms that never appeared under DSA often lands between 20% and 40%. Above 60% signals that negatives and conversion signal quality need attention.
How do I tell if AI Max is buying junk traffic?
Calculate CPA and conversion rate on new terms only — those absent from your pre-migration export. Meaningful spend with no conversions there is leakage, not expansion. Cut it with negatives rather than waiting for month-end.
What if CPA rises after migration?
A rise in the first three to five days is expected exploration. If it has not recovered by the end of week two, disable features one at a time — text customization first, then URL expansion — and see which one moves the number.
Can I migrate partially and keep some campaigns on DSA?
Until September 2026, yes, and that is the right approach: move in risk buckets. Once automatic upgrades begin the decision is made for you, so only the campaigns you moved and configured yourself will be set up the way you intended.
Does the migration affect conversion signal quality?
Not directly, but AI Max is more dependent on that signal. A single generic goal with no value differentiation gives expanded matching a poor target to learn from. Fix conversion values and offline import before you migrate, not after.
What if I have dozens of DSA campaigns?
Work in bulk: export the structure, mark risk buckets in a sheet, apply settings en masse. Walking through campaigns one at a time at that scale guarantees missed URL exclusions.
