Google's Forced AI Max Migration Creates a Measurement Crisis for Search Advertisers

Google began automatically upgrading eligible search campaigns to AI Max on September 1, a forced migration that rolls through the month with no opt-out for advertisers. Campaigns using campaign-level broad match or legacy Automatically Created Assets are being converted in place, while Dynamic Search Ads got a reprieve — their migration was pushed to February 2027 after advertiser pushback. Creation of new legacy campaign configurations was already blocked on August 3.

The migration is the most aggressive automation push Google has made in search advertising. AI Max bundles three capabilities that fundamentally change who controls a search campaign: automated search-term matching that decides which queries trigger an ad beyond the advertiser's keyword list, AI-generated headlines and ad copy, and final URL expansion that can redirect clicks to a different landing page than the one the advertiser specified. For measurement teams, the implications are immediate — when the platform controls the inputs (queries, copy, landing pages) and reports the outputs (conversions, ROAS), independent verification becomes both harder and more necessary.

The Numbers Don't Agree

Google says AI Max delivers an average of 7% more conversions at similar cost per acquisition when the full feature suite is active. That claim comes from the company's own aggregate data and excludes retail, the vertical where most AI Max adoption has occurred.

Independent measurement tells a different story. Smarter Ecommerce analyzed more than 250 retail search campaigns running AI Max and found a median 13% increase in conversion value — but at a median 16% higher CPA. The net ROAS impact was effectively zero at the median, and the range stretched from a 42% improvement to a 35% decline. The study, led by SMEC's Head of Ecommerce Insights Mike Ryan, has become the most-cited third-party benchmark in the PPC community precisely because it reveals what the average conceals: highly variable outcomes where a significant share of advertisers are worse off.

The gap between what the platform reports and what advertisers measure independently is now a 42-percentage-point spread in the worst cases. That gap matters because AI Max and Performance Max increasingly overlap on queries and channels, making last-click attribution unreliable — it overstates paid search one month and understates it the next, depending on which Google product claims the conversion.

Invalid Traffic Is Rising Faster Than Conversions

The measurement concerns extend beyond ROAS. Ad fraud detection firm Lunio published findings from 414 million retail ad clicks on August 12, revealing that AI Max campaigns carry 72% more invalid traffic than standard search campaigns running in the same accounts.

The trajectory is striking. AI Max invalid traffic rates in retail climbed from 2.46% in Q4 2025 to 5.28% in Q2 2026, more than doubling in nine months. Standard search campaigns moved in the opposite direction over the same period, falling from 3.72% to 3.07%. As Lunio's report put it, AI Max campaigns went "from the cleanest traffic in the channel to the dirtiest" in under a year.

The invalid traffic increase tracks directly to AI Max's broader query matching. When the system autonomously expands the queries that trigger an ad — matching searches the advertiser never bid on — it inevitably reaches lower-quality traffic sources. Advertisers paying for AI Max's broader reach are also paying for the bot traffic and misattributed clicks that come with it, a cost that Google's own reporting does not surface.

What the Migration Actually Changes

The September migration applies different defaults depending on what legacy feature a campaign was using. DSA users will eventually get all three AI Max features enabled: search-term matching, text customization, and final URL expansion. ACA users get matching and text customization. Campaign-level broad match users get search-term matching only.

For measurement teams, the critical change is the new search terms report. Google now labels expanded traffic with an "AI Max" match type and adds a Source column, providing the first visibility into which conversions came from AI-selected queries versus advertiser-specified keywords. This reporting upgrade was a direct response to advertiser demands, but it creates a new problem: maintaining clean A/B comparisons is nearly impossible when the platform can autonomously change the queries, copy, and landing pages within the test.

The forced migration also eliminates the control group. Before September, advertisers who kept legacy campaign structures had a natural baseline for measuring AI Max's incremental impact. With the migration complete, that baseline disappears. The only advertisers who retain it are those who proactively paused or restructured campaigns before September 1 — and even they face the challenge that Google's auction dynamics have shifted as the majority of the market moves to AI Max.

The Broader Pattern: Platforms as Black Boxes

Google's AI Max migration is part of a pattern that now spans every major ad platform. Meta's Advantage+ campaigns automate audience selection and creative. Amazon DSP's audience prediction models choose targeting without advertiser input. The Trade Desk's Kokai uses AI to optimize across channels. In each case, the platform absorbs decisions that advertisers and their measurement teams used to control.

The common thread is a measurement architecture that tilts toward the platform's self-reported numbers. When a platform controls targeting, creative, and landing page selection, its internal attribution model captures the full picture of what happened — but only from its own perspective. Incrementality testing, the method designed to answer whether a conversion would have happened without the ad, becomes harder to execute when the platform autonomously changes the variables that would normally be held constant in a test.

The IAB's State of Data 2026 report captured the resulting frustration: 60% to 75% of buy-side users said advanced measurement fell short on rigor, timeliness, trust, or efficiency. The September AI Max migration is likely to push those numbers higher.

What Measurement Teams Should Do Now

Audit AI Max traffic quality immediately. Use the new Source column in search terms reports to isolate AI Max-expanded queries. Compare conversion rates, bounce rates, and downstream revenue metrics for AI Max traffic versus keyword-matched traffic. If the quality gap is significant, use negative keyword lists aggressively to constrain the expansion.

Build measurement baselines before the window closes. If any campaigns have not yet been migrated, snapshot their performance data now. That pre-migration baseline is the only clean comparison point for evaluating AI Max's actual impact in your account. Once the migration is complete across your account, the opportunity for a natural control group is gone.

Move incrementality testing upstream. With AI Max and Performance Max overlapping on queries, last-click attribution is no longer a reliable signal for search spend allocation. Geo-based incrementality tests — using tools like Google's own Meridian GeoX or independent platforms — provide the causal evidence that platform-reported metrics cannot. The irony is not lost on the industry: Google's open-source incrementality tool may be the best defense against the measurement opacity created by Google's own campaign automation.

Negotiate for third-party verification access. Lunio's invalid traffic findings should be treated as a leading indicator, not an outlier. Demand contractual access to third-party traffic quality measurement on AI Max campaigns. If Google's automation is genuinely delivering incremental value, that value should survive independent scrutiny.

The forced migration to AI Max is not reversible, and for most advertisers the performance case will be ambiguous rather than clearly negative. The real risk is not that AI Max fails outright — it is that measurement teams lose the ability to know whether it is working. In an industry that has spent a decade building independent measurement infrastructure to check platform claims, the September migration is a test of whether that infrastructure can keep pace with the platforms it was designed to audit.