Two weeks after we covered TikTok's MCP server launch and a month after Meta's AI Connectors, Google quietly shipped the next move in the same direction. The headline news from the latest Google Ads update: Dynamic Search Ads are being upgraded to AI Max, and AI Max itself is expanding into Shopping, travel, and Performance Max via a new Search Themes beta.

If you run Google Ads at any scale, this is the most consequential change of the quarter. DSA campaigns won't disappear overnight, but the migration path is one-way, the new defaults are aggressive, and the surface area where Google's models decide what to show — and to whom — just got significantly larger.

This post breaks down what actually shipped, what changes operationally, the gotchas Google's announcement doesn't lead with, and what to do this week to stay in control.

What Google Announced

The May 2026 Google Ads update covers four moving parts. They're related but distinct:

Taken together, the message is consistent with what Meta and TikTok shipped in April and May: the manual ad-ops surface is shrinking, and the AI surface is becoming the default. The difference with Google is that it's happening to campaigns you already run, not just new ones.

What "DSA Becomes AI Max" Actually Changes

DSA has been around for a decade. It crawled your site, matched your pages to queries, and auto-generated headlines. Many advertisers ran it as a discovery layer alongside keyword campaigns. The AI Max successor keeps that core pattern and adds three things that change behavior in non-trivial ways.

1. Final URL expansion is on by default

Final URL expansion lets Google route a click to any URL on your domain it thinks better matches the query — not just the URL you specified. In DSA this was opt-in. In AI Max it's the default. If your site has thin pages, outdated content, or product pages that shouldn't get paid traffic, you need an exclusion list before flipping the switch, not after.

2. AI-generated headlines pull from on-page content

AI Max writes ad copy by reading the destination page. The good version: ads stay relevant as pages change. The bad version: a Friday afternoon content edit becomes a Monday morning ad headline you never approved. Branded language, regulated claims, and pricing references are the high-risk surfaces. Use the new text disclaimer controls if any of those apply.

3. Reporting consolidates but loses granularity

The DSA-style "search term + matched landing page" view collapses into AI Max's broader category buckets. You'll still see what queries triggered ads, but the line between "this is the search term" and "this is what Google inferred" gets fuzzier. Plan to lean harder on conversion-side analytics — your own GA4 or warehouse data — and less on the in-platform attribution view.

The Search Themes Beta for Performance Max

Search Themes is the most under-reported part of this update and arguably the most useful. Performance Max has been criticized since launch for being a black box: you feed in assets and audiences, Google decides everything else, and you can't tell the system "we also want to reach people searching for X."

Search Themes fixes that. You write short descriptions — a phrase, a concept, an intent — and Performance Max treats them as guidance for where to serve. Examples that work well:

These are not keywords. They're intent descriptions Google's model uses to expand its targeting into placements you may not have known to ask for. Early-access accounts have reported the biggest lift in industries with niche audiences where keyword-only PMax was underdelivering.

The caveat: Search Themes is guidance, not a hard constraint. Google still decides whether to honor the theme on any given auction. Treat it as a steering input, not a targeting filter.

Where to Push Back on the New Defaults

The launch coverage is uniformly positive. Here's the version with the marketing copy stripped out.

Final URL expansion still sends real money to the wrong pages

This is the single most common DSA complaint, and AI Max inherits it. The model has no idea that your /careers page shouldn't get paid traffic, that your /partners page is for partner referrals only, or that the blog post you wrote in 2022 has a CTA that no longer works. The first thing to do before the upgrade hits your account: build a URL exclusion list from your sitemap. Anything that isn't a conversion-capable landing page goes on it.

AI-generated headlines need a creative review checkpoint

"Set it and forget it" is the pitch. The reality is that AI-generated copy needs the same review cadence as human copy — just compressed into a weekly audit instead of a per-asset approval. Schedule a 30-minute review every Monday for the first month after migration. Pull the top 20 served headlines, flag anything off-brand, and add negative phrasing or excluded URLs as needed.

"AI Brief" is not a brief in the agency sense

The AI Brief tool lets you type a paragraph describing your messaging direction. It's helpful, but it's a hint to the model, not a contract. If you have non-negotiable claims — regulatory copy, exclusivity language, partner mentions — encode them as text disclaimers, not as AI Brief instructions. The disclaimer system guarantees presence; the AI Brief does not.

Cross-platform comparisons get harder, not easier

AI Max, Advantage+, and Smart+ all run on similar logic — model picks targeting, model picks creative, you pick guardrails. They do not expose comparable metrics. Be cautious about side-by-side ROAS comparisons across platforms after migration; you're often comparing different attribution windows, different conversion definitions, and different inclusion criteria for "incremental." If incrementality matters for your budget allocation, run real geo or holdout tests, don't trust the dashboards.

How AI Max Fits With Meta and TikTok's Direction

The three majors are converging on the same pattern. Each platform now ships an AI-first campaign type that takes minimal advertiser input and runs broad. The differences are starting to matter less than the similarities.

Google AI MaxMeta Advantage+TikTok Smart+
Headline statusDefault for new Search, replacing DSADefault for new campaigns in Ads ManagerPrimary surface for TikTok MCP server
Creative sourceReads landing pages, generates copyGenerates and tests variants automaticallySymphony AI generates video and avatars
Targeting inputSearch Themes, AI Brief, exclusionsBroad targeting, suggested categoriesSkills-based targeting, audience discovery
Reporting transparencyReduced vs. legacy SearchReduced vs. legacy ad setsSkill-level, not creative-level
Advertiser control surfaceDisclaimers, URL exclusions, briefsBrand-safety flags, exclusion listsPolicy via API, not UI

The strategic implication: the advertiser's job is shifting from execution to instrumentation. You no longer pick the bid, write the headline, or build the audience. You define what the model is allowed to do and measure whether it stayed inside the lines. That's a different skill set than the one most performance teams hired for in 2023.

What to Do This Week

If you run Google Ads, in priority order:

  1. Build a URL exclusion list from your sitemap. Identify every page that should never be a paid landing page (careers, partners, legal, outdated blog content, internal tools) and stage it before AI Max migration touches your account.
  2. Audit DSA campaigns for at-risk verticals. If you advertise in regulated industries — finance, health, legal — the AI-generated copy default is the highest-risk part of the upgrade. Plan disclaimer text now.
  3. Pilot Search Themes on one PMax campaign. Pick a niche-audience campaign that's underdelivering and write three to five themes. Compare a four-week window before vs. after. Real test, not vibes.
  4. Establish a weekly served-headline review. Pull the top 20 AI-generated headlines, mark anything off-brand, fix via exclusion or disclaimer. Calendar it.
  5. Decide your cross-platform measurement story. If you're going to defend budget shifts between AI Max, Advantage+, and Smart+, you need an attribution model that doesn't trust any of them blindly. Geo holdouts or MMM, not last-click in three different dashboards.

The Bigger Picture

Six months ago, the AI conversation in ad ops was about copilots — tools that helped media buyers do their work faster. That framing is over. AI Max, Advantage+, and Smart+ are not copilots. They're the campaign type. The buyer's job is to set the policy and watch the output.

That's not necessarily bad. The advertisers who win in this environment are the ones with the cleanest first-party data, the sharpest brand guardrails, and the most honest incrementality measurement. The advertisers who lose are the ones who treat AI defaults as "set and forget" and discover six weeks later that 30% of paid traffic was hitting the wrong pages.

The platforms shipped the automation. The leverage is in what you wrap around it.

Run AI Max, Advantage+, and Smart+ from one agent

Ads Agents adds the guardrails, exclusions, and weekly reviews these platforms don't ship by default — across Google, Meta, and TikTok in one place.

Get Started Free →