AI Max Now Lets You A/B Test Budgets Across Campaigns. Here's What to Measure (and What to Ignore).

Google claims advertisers using AI Max see roughly 14% more conversions on average. One practitioner test cited in industry coverage showed 145.5% higher spend for only 32.4% higher conversion value, with ROAS down 46%. Both numbers are real. Both are directionally useless without the right experiment design.

That gap between platform-reported lifts and actual pipeline impact is exactly why the new AI Max testing and planning tools matter for B2B teams. But the tools won't save you. Your measurement plan will.

What Google Actually Shipped

Starting in September 2026, advertisers can run multi-campaign A/B tests comparing different budgets and ROI/ROAS targets across multiple Search campaigns in a single experiment. Previously, AI Max experiments were limited to one-click toggles on individual campaigns. The new setup lets you test at the account level, which is where B2B SaaS teams need to be reading results anyway.

Two details worth noting. AI Max experiments now support brand and location controls, so you can run tests without stripping out ICP or geo guardrails. For teams with strict territory or vertical constraints, this removes a real blocker. Google also upgraded Performance Planner to forecast bidding and budget changes, then apply recommended edits directly into live campaigns.

A Gemini-powered "AI Brief" tool lets you define messaging, matching, and audience guidance in natural language. Google has stated AI Max is out of beta, with more than 500,000 advertisers using it. Eligible Search campaigns running campaign-level broad match or Automatically Created Assets will be auto-upgraded to AI Max starting September 1, 2026.

The Measurement Problem These Tools Don't Solve

Multi-campaign A/B tests are better than single-campaign toggles. But they still report on platform conversions. For B2B SaaS with sales cycles measured in weeks or months, platform conversions (form fills, demo requests) are leading indicators at best and vanity metrics at worst.

Controlled analyses cited in industry coverage argue AI Max's automated ad writing may shift demand between campaigns rather than create net-new demand. If Campaign A cannibalizes Campaign B's traffic and both report more conversions, your dashboard looks great while your pipeline stays flat. Account-level measurement is the only way to detect this.

One retail-focused report found AI Max campaigns saw 72% more invalid traffic than non-AI Max search campaigns in its dataset. Different vertical, different motion, and the sample matters. But it reinforces the point: validate traffic quality rather than assume efficiency gains.

Run It This Week: The Experiment Design That Actually Tells You Something

Setup: Pick 2–4 Search campaigns with enough volume to reach statistical significance within 2–3 weeks. Keep brand and location controls enabled. Set a clear holdout: one group runs AI Max with your proposed budget/ROAS change, the other holds current settings.

The hypothesis (make it falsifiable): If we increase budget by 20% on AI Max-enabled campaigns with tROAS held constant, then qualified pipeline (SQL or Opp) from paid search will increase by at least 10% within 30 days, because AI Max's broader matching will surface incremental high-intent queries we aren't currently covering.

Success metrics: Primary = SQLs or Opportunities sourced from paid search (CRM-attributed, not platform-reported). Secondary = cost per SQL, account-level conversion volume. Guardrail = cost per SQL doesn't exceed 130% of baseline. Stop-loss = if CPA rises more than 30% with no corresponding pipeline lift after two weeks, pause the experiment.

What to measure and what not to over-interpret: Ignore campaign-level conversion lifts in Google Ads. They'll almost certainly look good. Pull CRM data instead: how many of those conversions became qualified leads? How many reached opportunity stage? Use offline conversion imports or CRM feedback loops so Google's bidding learns what an SQL looks like, not just what a form fill looks like.

Before you enable anything: Audit brand exclusions, negative keyword lists, tracking templates, and Final URL expansion settings. AI Max will expand matching aggressively. If your negatives are stale or your URL routing is sloppy, you'll pay for irrelevant traffic before the experiment even reads out.

The Trade-Off You're Accepting

AI Max shifts the operational center of gravity from manual keyword control to signal quality and creative governance. Your competitive advantage moves from "better keyword lists" to "better conversion signals" and "tighter exclusion hygiene." Teams that feed Google clean pipeline data will get better results. Teams that optimize toward raw leads will get more raw leads.

The September 1 auto-upgrade deadline adds urgency. If you're running campaign-level broad match or ACA today, Google will move you to AI Max whether you've tested it or not. Better to run a controlled experiment now, with guardrails, than to wake up in September wondering why your CPA spiked and your pipeline didn't follow.

Google built better testing tools. That's genuinely useful. But the tools measure what Google can see. Your job is measuring what Google can't: whether those conversions turned into revenue.