AI Max for Search in 2026: What the Auto-Migration Actually Changes

On September 1, 2026, AI Max for Search stopped being optional: Google started automatically moving Search campaigns that use campaign-level broad match and legacy automatically created assets into AI Max. Dynamic Search Ads follow in February 2027. This is not a cosmetic setting change: it alters which queries trigger your ads, what headline the searcher reads, and which page they land on.

This guide covers exactly what flips, on which dates, what control you keep, and the checklist to run before the migration reaches your account. AI Max is not a disaster and it is not free money either — it rewards accounts with clean signals and punishes accounts with sloppy negatives faster than broad match ever did.

What AI Max actually is (and what it is not)

First thing to get straight: AI Max is not a new campaign type. Unlike Performance Max or Demand Gen, it is a feature set that switches on inside a regular Search campaign and replaces three older mechanics at once — campaign-level broad match, Dynamic Search Ads, and automatically created assets.

It has three moving parts:

• Search term matching. The system finds relevant queries beyond your keyword list, using your site content, your assets, and campaign history. Think of it as broad match and DSA merged into one model instead of two separate expansion engines.
• Text customization. Headlines and descriptions adapt to the specific query and landing page. Your written headlines stay in the pool; generated variants join them.
• Final URL expansion. Instead of a fixed destination, the system can route the click to a more relevant page on your site.

The practical difference from the old broad-match-plus-DSA setup: you used to manage two independent expansion sources with separate settings and separate reports. Now it is one layer with shared controls — brand inclusions and exclusions, URL exclusions, negative keywords — and unified reporting that labels how each search term was matched.

AI Max takes three levers away from you — query selection, ad copy, and landing page — and hands back one: the quality of the signals you feed it. Bad signals now produce bad results faster and at larger volume than before.

The migration calendar that matters

Google published a phased schedule. Treat these as announced windows — the exact notification date inside your account may land earlier or later within each window.

Date	What happens	What to do
August 3, 2026	Creation of new campaign-level broad match campaigns and legacy automatically created assets is closed everywhere	Already in effect. You cannot spin these up in the UI or via API
September 1–30, 2026	Automatic migration of existing campaign-level broad match and ACA campaigns into AI Max	Run the prep checklist before the switch. Capture baseline metrics
September 2026	In-account notifications encourage voluntary DSA upgrades	Upgrade voluntarily only inside a controlled experiment, not account-wide
January 15, 2027	Reminder notifications for advertisers still on legacy DSA	Last quiet window for testing
February 1–28, 2027	Automatic DSA migration; creating new DSA ad groups is removed permanently	By now your DSA logic should already live in AI Max deliberately
Around September 2027	Older Google Ads API versions supporting legacy entities sunset	Audit custom integrations, exports and scripts

One useful detail from the announcement: existing brand inclusions and exclusions carry over automatically. Your brand guardrails will not be wiped. Everything else you did not configure in advance, however, migrates exactly as it is.

What you still control versus what the model decides

The common mistake in automation debates is assuming control disappears entirely. It does not. There is meaningfully more control here than in Performance Max — the levers are just different ones.

You control	The model decides
Campaign and account-level negative keywords	Whether a specific query counts as relevant
Brand inclusions and exclusions	How the headline is phrased for that query
URL exclusions and landing page rules	Which allowed page to serve
Bid strategy and target CPA/ROAS	The bid in each individual auction
Geo, languages, schedule, observation audiences	Signal combinations inside your boundaries
Conversion action and its value	Who looks like a likely converter
Assets: headlines, descriptions, extensions	Which assets get assembled into the ad

The takeaway is blunt: preparing for AI Max is about 80% work on the left column. The tighter your fence of negatives, URL exclusions and correct conversion actions, the less room the model has to be expensively wrong.

The four-week prep checklist

Week 1 — Inventory and baseline

Start by finding what is actually affected:

1. Campaigns → filter by type “Search”. Flag the ones running campaign-level broad match.
2. Inside each Search campaign, check for Dynamic ad groups — they carry a distinct type label.
3. Ads & assets → find where automatically created assets are enabled.

Then capture a baseline. Pull the last 28 days for every affected campaign: impressions, clicks, CTR, CPC, conversions, cost per conversion, conversion value, impression share, and impression share lost to budget and to rank. Without this export you will not be able to separate the migration’s effect from seasonality a month later. For how to read the impression share family properly, see our breakdown of impression share and Auction Insights in Google Ads.

Week 2 — Build the negative fence

Expanded query matching sits on top of your keywords, so every gap in your negatives now costs more. Order of operations:

• Export 90 days of search terms and strip out every off-target cluster: “free”, “DIY”, “jobs”, “salary”, “how to become”, competitor brands you do not want, research-intent phrasing.
• Put them into account-level negative keyword lists, not into individual campaigns. Lists survive campaign rebuilds and copies; loose campaign negatives do not.
• Split lists by intent: “irrelevant intent”, “competitor brands”, “informational”, “careers and HR”. Selective attachment beats one giant list.

The method for building these lists is covered in detail in our guide to negative keywords and search terms — it applies directly to AI Max prep.

Configure brand exclusions separately. If you do not want expansion onto competitor brand queries — or onto your own brand, which you likely run in a dedicated campaign — that is set through the brand list at campaign level.

Week 3 — Landing pages and assets

Final URL expansion is the most underestimated part of AI Max. The system may send traffic to any indexable page on your domain it considers more relevant. Close off:

• Thank-you pages, carts, account areas, login screens.
• Blog and editorial sections, if this is a performance campaign rather than a content funnel.
• Expired promotions and legacy landers still reachable by URL.
• Careers pages, legal documents, broken filter results.

Use campaign-level URL exclusions, and exclude by directory pattern rather than page by page — otherwise every new page becomes a new leak.

Then the assets. Text customization does not invent positioning; it recombines what already exists in your ads and on your site. Three tired headlines about “best prices” produce three tired generated variants. Build 12–15 headlines and 4 descriptions per ad group covering distinct angles: price, speed, guarantee, range, geography, service. Our guide to ad extensions and assets in Google Ads covers the mechanics.

Week 4 — Conversions, bidding, and one experiment

AI Max hunts for new demand more aggressively, which amplifies any distortion in your conversion accounting. Check three things:

1. What counts as a conversion. If your Primary conversion actions mix “viewed contact page” and “clicked phone” with actual purchases, the algorithm optimizes toward the cheapest and most frequent one. Keep only actions you would happily pay for.
2. Whether conversions carry value. Even crude tiers (cheap product lead = 1, enterprise lead = 5) give bidding a signal it previously lacked.
3. Signal quality. If half your conversions die in the browser, no amount of automation fixes it. That is a topic in itself — see our guide to server-side tagging with sGTM for paid traffic.

On bidding: AI Max pairs badly with Maximize Clicks. Broader matching plus a strategy that ignores conversions is the fastest route to more traffic and less revenue. Run tCPA or tROAS; the trade-offs are in our breakdown of Google Ads bidding strategies.

Most importantly, do not wait passively for auto-migration. Launch a campaign experiment: a copy with AI Max enabled against your current setup, 50/50 traffic split, running three to four weeks or until each arm has at least 30–50 conversions. When the forced switch arrives, you will have your own numbers instead of forum opinions.

Measuring the migration honestly

The classic error: compare CPA for the month before and the month after, see a 12% increase, and blame AI Max. In that same month, seasonality, competitive pressure, your pricing, and your inventory all moved too.

Do this instead:

• Compare experiment arms, not time periods. Same window, same demand, split traffic.
• Look at total conversion value at a fixed budget rather than CPA in isolation. AI Max frequently delivers more conversions at a slightly higher CPA — if your unit economics absorb that, it is a win. How to calculate your ceiling: media buying unit economics.
• Test incrementality separately. Some of the “new” query volume may cannibalize your own organic or brand traffic. The clean method is a geo holdout, described in incrementality and geo experiments.
• Read the search terms report weekly with the new match-type segmentation. For the first two to three weeks this is a scheduled ritual, not a “when I get around to it” task.

Seven expensive mistakes on the way in

1. Migrating during peak season

Any targeting expansion triggers a learning period. If your peak is September and auto-migration lands in September, move key campaigns manually in August inside a controlled experiment so the learning happens on calm traffic.

2. Negatives only at campaign level

Loose campaign negatives vanish when campaigns get rebuilt or duplicated. Account-level lists do not. That is the difference between configuring once and re-cleaning every month.

3. Brand and non-brand in the same campaign

If brand queries live in your generic campaign, AI Max will happily shift budget toward cheap brand traffic and report a beautiful CPA. Revenue will not move. Keep brand separate and exclude it elsewhere.

4. Optimizing toward a micro-conversion

“Scrolled 30 seconds” or “clicked a button” sitting in Primary conversions turns expanded matching into a machine for producing cheap, useless actions.

5. An open site with no URL exclusions

Especially painful for content-heavy sites. Traffic drains into blog posts, conversions do not follow, and the report reads like “bad queries” when the real fault is an open door.

6. No baseline

Without a “before” export you cannot prove to yourself or a client that anything improved. Fifteen minutes of work, one-time cost, permanent value.

7. Ignoring the API deadline

Custom exports, automated rules via scripts, BI dashboards on the Google Ads API — legacy API versions supporting deprecated entities are being retired. Check this before your Monday report silently comes back empty. What is worth automating and how: Google Ads scripts and automated rules.

Who benefits and who should be careful

Situation	Likely outcome	What to reinforce
Wide catalog, large site, many landing pages	Upside: there is room to expand into	URL exclusions, page feed, page quality
Narrow niche, 3–5 landing pages, stable demand	Neutral to negative: little to expand into, more noise	Hard negatives, tight geo, query review
Rich conversion data (100+ per month)	Upside: the model has something to learn from	Conversion values, offline data
Thin conversion data (under 20–30 per month)	Risk: slow learning, noise eats budget	Supporting micro-conversions, longer test window
Lead gen with manual qualification	Risk of a rising junk-lead share	Offline conversion import reflecting real lead quality

One more category deserves attention: accounts with limited history and hard requirements for uninterrupted delivery. Automation loves data volume and predictable budgets, so a young account with a swinging budget will show a muddy AI Max result. Teams that need reliable, seasoned Google Ads infrastructure can look at agency Google Ads accounts from PPC Rebels — spend history included, no cold-start surprises.

What to do tonight

1. Export the list of Search campaigns with campaign-level broad match, DSA ad groups, and automatically created assets.
2. Capture a 28-day baseline for each and save it as a separate file.
3. Build two account-level negative lists: “irrelevant intent” and “competitor brands”.
4. Set URL exclusions for the site sections that should never receive paid traffic.
5. Open the Conversions section and leave only revenue-relevant actions marked Primary.
6. Launch a 50/50 experiment on one mid-sized campaign. Not the largest, not the smallest.
7. Put a weekly search-terms review in your calendar for the next month.

Seven steps turn the migration into a managed change rather than a Monday surprise. And one final note: if performance dips after the switch, do not roll everything back on day one. Give the system 14 days at a stable budget — most week-one “disasters” flatten out by week two.

FAQ: AI Max for Search campaigns

Can I opt out of AI Max entirely?

Not from the migration of legacy entities: new campaign-level broad match campaigns and new DSA ad groups can no longer be created, and existing ones are moved automatically. Inside a Search campaign, however, individual components are configurable, and you set the boundaries through negatives, brand exclusions, and URL exclusions.

Is AI Max the same as Performance Max?

No. Performance Max is a separate campaign type covering Search, Display, YouTube, Discover, Gmail and Maps. AI Max lives inside a Search campaign and touches Search inventory only, while keeping your keywords, your search terms report, and familiar controls.

Do keywords still matter?

Yes. Keywords remain the primary signal, with expanded matching layered on top. Semantics quality still counts, and match types are not going away.

What happens to my existing DSA campaigns?

Per the announced schedule: voluntary upgrade prompts appear in September 2026, reminders in January 2027, and automatic migration runs through February 2027, after which new DSA ad groups can no longer be created.

Are brand settings preserved?

According to Google’s announcement, existing brand inclusions and exclusions carry over automatically during migration. Verify them manually afterward anyway — it takes two minutes.

How do I see which queries came from expansion?

The search terms report includes match-type segmentation that separates expanded matches from keyword matches. This is your main control instrument in the first weeks.

How long is the learning period?

Plan for 7–14 days at a stable budget and bid strategy. Changing targets, budgets or structure mid-way restarts learning, and you will never see clean numbers.

Should I enable AI Max on brand campaigns?

Usually not. Brand demand is narrow and well covered by exact match; expansion here mostly blurs reporting and pulls in off-target traffic. Keep brand in its own tightly fenced campaign.

How do I prove AI Max is profitable rather than just spending more?

Compare experiment arms within the same period, and judge on total conversion value at a fixed budget rather than CPA alone. For true contribution, run a geo incrementality test.

Irrelevant queries spiked after migration — what now?

In order: account-level negative lists, URL exclusions, then a review of which action is set as your Primary conversion. Nine times out of ten the cause sits in one of those three places, not in the automation itself.

Does AI Max affect Quality Score?

Quality Score is still calculated per keyword. But ad relevance and landing page experience matter more than before, because both the copy and the destination are now chosen dynamically — a weak site drags down more combinations.

Do I need to change anything in my API integrations?

Yes, if you run custom exports or scripts. API versions supporting deprecated entities are scheduled to sunset around September 2027. Check which version your reporting depends on well before that.