If your DSA program is a landing-page governance system (not “just” a search tactic), AI Max is a control change disguised as an upgrade. The forced migration starts September 2026—so the practical question isn’t “is AI Max good?” It’s: what guardrails can be made real before Google makes the switch for you.

If your DSA program is basically a landing-page governance system (not “just” a search tactic), AI Max is a control change disguised as an upgrade. The forced migration starts September 2026—so the practical question isn’t “is AI Max good?” It’s: what guardrails can be made real before Google makes the switch for you.

Google says AI Max for Search is now generally available globally and will replace Dynamic Search Ads (DSA). The timeline is tight: a voluntary migration window runs April–August 2026, then remaining eligible DSA campaigns begin automatic migration in September 2026, with no opt-out. New DSA creation also gets disabled across the Google Ads UI, Google Ads Editor, and the Ads API in that September phase. That’s not a product tweak. That’s an operating model change. (Sources: Google Blog; Deeper Insights)

And yes, Google is dangling performance claims—+7% average conversion lift (or conversion value) at similar CPA/ROAS when all AI Max features are enabled, plus other cited lift ranges (+14% average broader conversion lift; up to +27% improvement for campaigns that relied on exact and phrase match). But those numbers don’t answer the thing demand gen leaders actually lose sleep over: where traffic lands, what it says, and whether “more conversions” turns into qualified pipeline. (Source: Deeper Insights)

What changed: DSA was a structure; AI Max is a system

DSA’s superpower wasn’t magic. It was structure. Many teams used URL/path-based campaign structuring to enforce landing page governance at scale—especially on large sites where “the wrong page” isn’t a minor issue, it’s a pipeline leak.

AI Max shifts the center of gravity to three capabilities that are explicitly more automated: Smart Search Term Matching (using intent/session/behavioral signals), Dynamic Text Customization (AI-generated headlines and descriptions), and Final URL Expansion (routing to pages Google thinks are relevant across the site). If that sounds like “more reach,” it is. It also sounds like “less predictable routing,” because it is. (Source: Deeper Insights)

This is where the control debate got loud. In Search Engine Land’s coverage, advertiser Gabriele Benedetti argued that AI Max doesn’t replicate DSA’s URL/path-based structuring—which is exactly the kind of thing that’s invisible until it breaks. (Source: Search Engine Land)

Google Ads Liaison Ginny Marvin responded with the official counterpoint: AI Max supports URL rules/combinations, page feeds with custom labels, and URL inclusions/exclusions at the ad group and campaign level. She also noted that when migrating from DSA, unsupported rules can remain active as read-only—still functioning, but no longer editable. Google has also said it plans to expand controls further, including content and title-based exclusions at the account level later in 2026. (Source: Search Engine Land)

The nut graf: the risk isn’t “automation.” It’s unmeasured routing

For B2B SaaS, the downside case isn’t a messy keyword list. It’s Final URL Expansion quietly sending high-intent searches to pages that convert worse to demos, trials, or contact sales—blog posts, help docs, partner pages, whatever happens to look “relevant” to the model. That can raise conversion counts while lowering qualified pipeline rate. Clean dashboard. Messy business.

That risk is amplified by the timeline. Digital Applied compared the April–September 2026 window (~5.5 months) to the Smart Shopping → Performance Max runway in 2022 (~9 months), calling this about a 40% shorter migration period. If a team manages a big DSA portfolio, “we’ll get to it” is how forced migrations turn into a quarter-long attribution argument. (Source: Digital Applied)

And results won’t be uniform. Digital Applied’s agency-reported benchmarks across 250+ campaigns show median revenue up +13% and median CPA up +16%, but only 22% hit original ROAS targets, with ROAS ranging from −35% to +42%. That spread is the whole story: the right posture is controlled testing with guardrails, not belief. (Source: Digital Applied)

If you only change one thing, change this: build a routing holdout

The primary tactic is simple to say and annoying to implement: treat landing-page routing as the experiment variable, not “AI Max vs not AI Max.” Because if routing goes sideways, everything downstream (lead quality, sales cycle, CAC payback) goes sideways with it.

Here’s the falsifiable hypothesis—make it specific enough that the readout can’t wiggle out of accountability:

The hypothesis (make it falsifiable): If AI Max is launched with strict URL governance (URL rules/inclusions/exclusions and page feeds with labels), then qualified pipeline per click will stay within a defined band versus the DSA baseline, because high-intent queries will be routed to the same high-converting commercial pages rather than mixed-intent content.

Two details matter here. First: “qualified pipeline per click” (or per impression, depending on volume) is the point—don’t let a conversion-lift headline replace a funnel metric. Second: the mechanism is routing. Not the ad copy. Not the match logic. Routing.

Digital Applied also flagged a hygiene problem that should make any operator pause: in one case study, 99% of impressions generated zero conversions when they analyzed roughly 30,000 AI Max search terms. That’s not proof AI Max is bad. It’s proof you need monitoring and exclusions, because waste shows up fast when matching expands. (Source: Digital Applied)

Run it this week: a migration-safe test plan (April–August 2026)

Here’s the 5-minute version you can run this week. The goal isn’t perfection; it’s to avoid being surprised in September 2026 when auto-migration starts and new DSA creation is disabled in the UI, Editor, and API. (Sources: Deeper Insights; Search Engine Land)

Success = qualified pipeline rate holds (or improves) while spend stays within plan. Guardrails = lead-to-SQL rate and sales-accepted rate don’t drop beyond an agreed band. Stop-loss = if CPA rises past a threshold and lead quality drops for two consecutive weeks, tighten URL governance and query exclusions before scaling.

One more operational gotcha: Digital Applied reported a +40% higher success rate when all AI Max features were enabled, plus Quality Score improving from 6.8 to 7.3 in their tests (23 tests; 16 mature accounts). That’s the trade-off in numbers: sometimes the “more control” configuration is also the “less upside” configuration. The only sane approach is to decide, in advance, what upside is worth what risk. (Source: Digital Applied)

The trade-off: control isn’t gone. It’s relocated.

There’s a temptation to frame this as “Google took away control.” The more accurate framing is harsher: Google moved control from the campaign’s visible structure (DSA URL paths) into a set of rules, labels, and exclusions that must be actively maintained—while matching and creative get more automated. That’s a different job.

When this is wrong: if the site is small, the page taxonomy is clean, and the conversion event is tight (and defended), AI Max’s automation may outperform the DSA-era approach without much drama. Google’s own cited lifts (+7% at similar CPA/ROAS when all features are enabled) are plausible in accounts where intent coverage was constrained.