Est. reading time: 8 minutes

Most Search accounts have an AI Max prompt sitting in them that has been safe to ignore for over a year. It was a beta, and waiting cost nothing. Google closed that window in April 2026 by moving AI Max out of beta and attaching a retirement schedule to three features plenty of accounts still depend on. Dynamic Search Ads, automatically created assets, and the campaign-level broad match setting are all being folded into it.

That reframes the decision. Campaigns still running those legacy features will be upgraded automatically, with settings Google configures to mirror the old setup. What remains to decide is whether you set up AI Max deliberately or let the auto-migration set it up for you. Either way, it helps to know that the toggle turns on more than one thing.

AI Max for Search is three features behind one toggle

AI Max for Search is a settings layer inside standard Search campaigns, not a new campaign type. Structure, budgets, bidding, and keywords all stay where they are. What changes is how the campaign matches queries, what the ads say, and which page the click lands on. Those jobs belong to three separate features with three separate risk profiles, laid out in Google’s feature documentation for AI Max:

• Search term matching expands delivery past your keyword list using broad match and keywordless technology. The system reads your keywords, ad copy, and URLs, then matches into relevant queries you never bid on. It defaults to on with AI Max and can be switched off per ad group.
• Text customization, formerly called automatically created assets, uses your existing ads, landing page copy, and generative AI to write new headlines and descriptions shaped to the query. It is controlled at the campaign level.
• Final URL expansion lets Google send the click to whichever page on your domain it predicts will perform best for that query, limited to pages relevant to the search and themed to the ad group. It is also campaign level, and it requires text customization to stay on so the generated copy matches the chosen page.

Two operational details deserve attention before anything gets enabled. When final URL expansion selects a different landing page, pinned assets in your responsive search ads are ignored, so a pinning setup built for compliance or message control does not survive the feature intact. And Google flags that AI Max adds little in campaigns limited by budget, because query expansion needs headroom to matter.

The toggle comes with a retirement schedule

The reason this stopped being optional homework is the migration behind it. Google is retiring three legacy features into AI Max, and the announced timeline now carries two dates. Campaigns using automatically created assets or the campaign-level broad match setting begin auto-upgrading in September 2026. Dynamic Search Ads received an extension in June, when Google pushed the DSA sunset and auto-migration from September 2026 to February 2027, reopened DSA creation as of June 15, and set that window to close again in January 2027.

What the auto-upgrade enables matters more than when it runs, and Google has published the default mapping. DSA campaigns come over with all three features on and their dynamic ad groups converted to standard ad groups, a reasonable translation, since DSA already generated headlines and picked landing pages. Broad match setting users get search term matching alone, which quietly adds keywordless matching they never ran. Automatically created assets users get text customization plus search term matching, so an account that opted into AI-written headlines will also start entering auctions its keywords never touched.

The fix is to run the upgrade on your own schedule. Google provides upgrade tools that port DSA settings and history into standard ad groups, and one-click experiments that test AI Max against the current setup before a whole campaign commits. A controlled migration in a quiet period beats discovering the new defaults mid-quarter, which is the logic behind the extension itself. Google moved the date to keep forced migrations out of Q4.

Keywords become one input among several

The deepest change in practice is what happens to the keyword list. With search term matching on, keywords stop defining the outer edge of where the campaign serves. They become source material, alongside your ad copy and landing pages, that the system reads to decide which queries to enter. This formalizes a shift that has been underway for years, and it is part of why the old Google Ads playbook keeps getting more expensive. Google has also positioned AI Max as one of the main ways Search ads surface in AI Overviews and AI Mode, which explains the pressure behind the rollout.

The precision you used to get from tight match types moves into a different set of controls. Locations of interest brings geographic intent filtering down to the ad group level, a control that only exists inside AI Max, and it applies even on keywordless matches. Brand inclusions restrict serving to searches involving brands you specify, at the campaign or ad group level, while brand exclusions keep you out of searches around brands you name at the campaign level. Negative keywords keep working, and they remain the fastest way to shut down bad expansion.

The reporting supplies the audit trail. The search terms report adds an AI Max match type and a source column showing whether a match came from broad match expansion or keywordless matching, plus a combined view tying each search term to the headline and URL that served with it. The keywords report adds AI Max summary rows, and the landing pages report adds a Selected by column that isolates the pages final URL expansion chose. Expansion without a weekly read of these views is how budgets drift, the same discipline problem we covered in how to read Google Ads data without fooling yourself.

Asset optimization decides what the ad says and where the click lands

Text customization and final URL expansion are the half of the suite that touches the customer directly, and they deserve the most guardrails. Generated copy is built from your ads, your landing pages, and your assets, so the quality of what the model writes is capped by the quality of what it reads. The pattern we see on ecommerce sites is stale input (expired promotions still live on landing pages, thin product descriptions, blog posts written to rank). All of it is eligible source material. Text guidelines let you set instructions the generated copy must follow, and asset removals let you delete anything that slips through, but the higher-value work is cleaning the source before the model reads it.

Final URL expansion applies the same logic to landing pages. Google routes each click to the page it predicts will convert for that query, and your control over the eligible set comes from URL exclusions at the campaign level and URL inclusions at the ad group level. Build the exclusion list before the feature turns on (careers pages, policy pages, out-of-season collections, anything paid traffic should never hit). Two failure modes are worth checking in the same pass. Tracking templates that handle the lpurl tag in nonstandard ways can send AI-selected landing pages to 404s, and a page that wins the relevance prediction can still lose the visitor after arrival, a conversion problem no amount of matching fixes.

Turn it on as a test, with the guardrails already set

Google’s own case for the full suite is an average of 7% more conversions or conversion value at a similar CPA or ROAS, compared with running search term matching alone, from its 2026 internal data for non-retail advertisers. Treat the number as directional. It is an average across accounts that opted in, measured by the party selling the upgrade, and your account will produce its own answer.

The rollout that keeps you in control runs in a specific order. Guardrails come first (brand exclusions, URL exclusions, negatives, text guidelines), because every one of them is cheaper to configure before matching starts than after a bad query has spent money. Then enable AI Max in one or two campaigns with clean conversion tracking and enough volume to read, using a one-click experiment so the original setup keeps serving as the baseline. Judge the result on conversions or conversion value at your target CPA or ROAS over a full purchase cycle, not on the reach it adds. Reach is the one outcome the feature guarantees.

There are also legitimate reasons to run the suite partially. Regulated categories and brands with strict claim language can run search term matching with text customization off, since the settings are separate. Accounts with thin or messy conversion data should fix tracking before enabling any of it, because everything in the suite optimizes toward whatever signal it is given, including the wrong one.

AI Max is the rare Google rollout where the deadline eventually makes the decision for anyone who declines to. Between now and the auto-upgrades, every account running DSA, automatically created assets, or the broad match setting has a window to move over with guardrails chosen and a test already read. We would spend the window. And if you want us in the account for it, start here.