Google AI Max Replaces DSA With a 13% Revenue Lift (and 16% Higher CPA)

Google officially moved AI Max for Search out of beta on April 15, 2026, and set a September 2026 deadline for mandatory migration from Dynamic Search Ads. An independent 250-campaign study found AI Max delivers a median 13% revenue increase alongside 16% higher CPA, with up to 63% of "new" traffic cannibalizing queries existing keywords already covered. How you configure the migration in the next five months determines whether you gain or lose efficiency.

Most accounts already left DSA behind

DSA has been on life support for a while. David Dweck at Go Fish Digital told Digiday that client DSA spending was already "nonexistent" before the announcement. If you're running a modern search account with broad match and Performance Max, DSA was probably redundant months ago. Google just made it official.

What actually changed: AI Max consolidates Dynamic Search Ads, Automatically Created Assets, and campaign-level broad match into a single AI layer on top of your existing search campaigns. It's not a new campaign type. It's Google's AI taking over the targeting decisions DSA used to handle, plus the creative generation ACA handled, plus expanded URL matching. All in one toggle.

The voluntary upgrade window opened today. September 2026 is the hard cutoff, after which you can't create new DSA campaigns and existing ones get auto-migrated. Seven marketers interviewed by Digiday showed minimal pushback to the transition, which tells you something about how little DSA was being used in practice.

Google's numbers versus the numbers that aren't Google's

Google's official pitch: AI Max delivers 14% more conversions at similar CPA/ROAS, rising to 27% for campaigns primarily using exact and phrase match. L'Oreal reportedly saw 2x higher conversion rates at 31% lower cost per conversion. Those numbers are real. They're also Google's best cases.

Mike Ryan at Smarter Ecommerce analyzed 250+ campaigns independently and the picture is messier:

• Median revenue increase: +13%
• Median CPA increase: +16%
• ROAS range across accounts: +42% to -35%

That ROAS range is the number I keep coming back to. Some accounts saw real gains. Others lost more than a third of their return on ad spend. The median advertiser is paying 16% more per acquisition for a 13% revenue bump, which means margins get tighter even when the top line grows.

It's like hiring a second sales rep who closes 13% more deals but mostly by poaching leads your first rep was already working. The revenue line looks good on a dashboard. The unit economics tell a different story.

One detail Ryan flagged that Google's pitch "conspicuously excludes": retail. For ecommerce advertisers, the data set is thinner and the conclusions less clear. If your business runs on shopping campaigns, the 14% uplift number was not built with you in mind.

63% of the "new" traffic is not new

This is probably the most uncomfortable finding from the independent research. In campaigns already running broad match, up to 63% of traffic AI Max claimed as incremental was recycled from queries the account's existing keywords already covered.

It wasn't finding new searchers. It was taking credit for traffic you were already getting.

One account saw 69% of its search impressions consumed by competitor brand terms. Another generated 500,000 monthly impressions on Google's Search Partner Network with a 0.07% conversion rate, compared to 3.04% on standard Google Search. If you're wondering whether that second number is a typo, it is not. That's real money spent on traffic that converts at roughly 1/43rd the rate of your normal search traffic.

I don't think this means AI Max is fundamentally broken. It means the default settings are aggressive, and if you flip it on without configuring the controls, it will happily spend your budget on queries you probably don't want. Which is, honestly, kind of the pattern with every Google automation product since Performance Max launched.

The controls exist (most people just don't configure them)

What separates AI Max from the earlier Performance Max frustration is that Google actually built in controls this time. They're not obvious if you don't go looking, but they're there. This is a genuine improvement, and it's worth acknowledging even while being skeptical about the defaults.

What you can actually configure:

• Text guidelines define messaging boundaries for AI-generated headlines and descriptions. These shipped globally in early 2026 and act as guardrails for what the AI can write.
• Brand controls at campaign and ad group level let you block competitor terms or restrict to your own brand queries.
• Final URL expansion can be toggled off. If you don't want Google routing traffic to pages you didn't choose, turn this off on day one.
• Search Partner Network can be disabled. Based on the 0.07% conversion rate data above, you probably should.
• Query-level reporting is available and detailed, which is a genuine step up from Performance Max's opaque reporting.

The difference between an AI Max migration that works and one that bleeds budget is roughly 30 minutes of configuration. Set the text guidelines. Turn off Search Partners. Review the query report weekly for the first month. That is the entire playbook.

Running the migration without losing visibility

If you're still on DSA, or if you've been ignoring the AI Max toggle in your account, here's what I'd do before September:

Audit your DSA query reports now. Pull 90 days of search term data and identify which queries actually convert and at what CPA. Those are the queries you need to verify AI Max covers when DSA disappears.

Layer AI Max onto existing search campaigns first. Don't build a new campaign. Turn on AI Max features (text customization, search term matching) inside your current campaigns and run both in parallel for 2-4 weeks. This gives you a clean comparison without blowing up your account structure.

Kill Search Partner Network immediately. The conversion rate data from the 250-campaign study is pretty damning. Unless your account is the rare exception, SPN is inflating impressions without delivering meaningful conversions.

Set text guidelines before enabling text customization. The AI will generate headlines and descriptions on its own. If you don't define what your brand sounds like, you'll get generic copy that could belong to any competitor in your space.

Watch the cannibalization metric. Compare total search impressions and conversions before and after enabling AI Max. If impressions jump but conversions stay flat, the AI is mostly claiming existing traffic rather than finding new traffic. That tells you to tighten URL expansion and keyword controls. (We covered a similar pattern with AI ad agents overpromising on automation while underdelivering on precision.)

Tucker Matheson at Markacy expects "brands will likely see higher reported CVR but have less clarity on drivers." That's a polite way of saying the dashboard will look better while the underlying economics get murkier. Query-level reporting is your counter to that. Use it weekly, not monthly.

Five months is more time than you think

Google has been making this move for years. Broad match, then Performance Max, now AI Max. Each step follows the same arc: Google takes control of targeting, reports aggregate uplift, and leaves individual advertisers to figure out the edge cases on their own. The trend line is not reversing because advertisers on Reddit are unhappy about it.

From what I've seen, the accounts that come out of these forced migrations in the best shape aren't the ones that resist. They're the ones that adopt early, while support channels aren't overwhelmed and before the documentation calcifies into FAQ pages answering questions nobody actually asked.

The first two weeks of configuration matter more than the last four months of monitoring. Set it up now, review data weekly through May, and by June you'll have a clear read on whether your account falls in the +42% ROAS group or the -35% one. Either way, you'll know before September makes the decision for you. That's worth 30 minutes this week.

By Notice Me Senpai Editorial