Google's Ads Liaison Ginny Marvin spent the week after Google Marketing Live 2026 answering questions that most advertisers were afraid to ask in public. The answers matter because AI Max is no longer optional. By September 2026, Google will force-migrate all remaining Dynamic Search Ads, Automatically Created Assets, and broad match campaigns into AI Max. The question is not whether to adopt it, but how to adopt it without wrecking your CAC payback.
The core shift is architectural. Marvin explained in a recent deep-dive Q&A that the integration of Gemini into Google's ads quality stack has fundamentally changed how queries are processed. The system now matches ads based on inferred user intent rather than keyword syntax. For B2B marketers who spent years building elaborate keyword architectures, this is the equivalent of being told your spreadsheet is now a suggestion box.
The Volume Trap
Google's published numbers look compelling: advertisers activating AI Max typically see 14% more conversions at similar CPA/ROAS, with campaigns using mostly exact and phrase match seeing 27% uplift. But those numbers measure conversions, not qualified pipeline.
One practitioner ran a 14-day head-to-head test and found AI Max produced 41% more form fills but 73% fewer qualified leads. The cost per qualified lead jumped from $178 to $313. The dashboard looked great. The sales team had to filter through twice as much noise to find the same number of real prospects.
This is the math your CFO needs to see before you flip the switch. AI Max optimizes for the conversion action you define. If that action is a form fill, you will get more form fills. If your sales team then spends 40% more time qualifying garbage, you have not improved efficiency. You have shifted cost from media to labor.
What Marvin Actually Said About Controls
The good news is that Google is building in controls that make AI Max usable for teams that care about lead quality. In her post-GML Q&A session, Marvin clarified that AI Brief allows advertisers to set both negatives (what not to focus on) and positives (what to focus on). You can say things like "never mention prices" in messaging guidelines. The system will show sample assets and searches so you can edit and refine before proceeding.
Three types of guidelines matter:
- Messaging Guidelines: Tell AI Brief exactly what ads should or shouldn't say.
- Matching Guidelines: Create search query boundaries for the types of searches you want to show up for, or avoid.
- Audience Guidelines: Describe the type of consumer you're targeting to serve more tailored messages.
AI Brief is rolling out in English for AI Max Search campaigns over the coming months, followed by Performance Max and AI Max for Shopping. If you are running regulated campaigns, text disclaimers now guarantee required legal text always appears in your ads, even when using Final URL Expansion. This solves a real problem: advertisers in financial services, healthcare, and legal previously had to choose between compliance and AI optimization.
The B2B-Specific Problem
B2B lead generation has a structural mismatch with AI Max's default behavior. Buyers do over 61% of research before contacting a vendor, and in 95% of deals, the winner was already on their shortlist. AI Max's keywordless expansion tends to capture top-of-funnel and educational queries that your keyword list would never have matched. For B2C, that is often fine. For B2B, it means paying for clicks from people who are 18 months away from a purchase decision.
The keywordless expansion only works when your landing page content is clear and specific. If your pages lack clear entity definitions, consistent messaging, or verifiable claims, the AI will generate weak ad variations and match you to low-intent queries. This is not a Google problem. It is a content infrastructure problem that AI Max exposes.
Current B2B Google Ads benchmarks show exact match keywords deliver 2x better cost per MQL than phrase match ($1,200 vs $2,800). Broad match, which AI Max enables by default, runs even higher. The recommendation from practitioners is to start with exact match on high-intent terms before expanding, but AI Max inverts that logic by expanding first.
The Pilot Design That Protects Your Budget
Before enabling AI Max across your account, run a controlled test with a sales-team feedback loop. The structure that works:
- Isolate one campaign with sufficient volume (minimum 30 conversions per month).
- Run AI Max in that campaign while keeping a control campaign with your existing keyword structure.
- Track not just form fills but CRM-marked qualified leads.
- Measure cost per qualified lead, not cost per conversion.
- Run for at least 14 days, ideally 30.
Marvin confirmed there are no plans to deprecate Search campaigns. AI Max is intentionally designed as an optional suite of features you can enable rather than a new campaign type or migration. This means you have time to test properly.
The brands seeing real results are the ones feeding CRM data back to Google. Enhanced conversions for leads now pass hashed email data from form submissions to Google's algorithm, and value-based bidding lets you assign different values to MQLs vs SQLs vs closed deals. If you are optimizing for form fills instead of pipeline quality, you are training the algorithm to find the wrong people.
What to Do This Week
The September 2026 migration deadline is real. If you wait until August, you will be scrambling. Start now with three actions:
First, audit your landing page content. AI Max pulls signals from your pages to determine query matching. If your pages are generic, your matches will be generic. Make sure each page has clear entity definitions, specific use cases, and language that distinguishes your ICP from everyone else.
Second, set up offline conversion tracking if you have not already. Connect your CRM so Google learns which leads actually progress through your funnel. Without this, AI Max will optimize for volume, not quality.
Third, run a controlled pilot with the structure above. Document your cost per qualified lead in both conditions. Bring that data to your next pipeline review. If AI Max improves qualified pipeline at similar or better cost, expand. If it does not, you have evidence for why you are using the controls to constrain it.
The shift from keywords to intent is not reversible. The question is whether you control the transition or let it control you.
