Google's AI Max for Search campaigns promises a 14% to 27% conversion lift with a single toggle. That claim, published in Google's May 2025 announcement, has been circulating through every paid media Slack channel and board deck for the past year. What hasn't circulated as widely: independent testing by Monks Agency found that 99% of AI Max impressions generated zero conversions across roughly 30,000 search terms, and a LinkedIn poll of PPC professionals showed only 16% reporting good performance.
That gap between vendor promise and operator reality is exactly the kind of math your CFO will ask about. So let's model it.
What AI Max Actually Does
AI Max is not a new campaign type. It's an opt-in feature layer you enable within existing Search campaigns. According to Google's documentation, it bundles three capabilities under one toggle:
Search term matching expands your keyword targeting using broad match and what Google calls "keywordless" technology. The system learns from your existing keywords, ad copy, and landing pages to serve ads on queries you never explicitly bid on. Think of it as Google saying, "We'll find the searches you missed." The risk is obvious: Google's definition of "relevant" may not match yours.
Text customization (formerly Automatically Created Assets) generates headlines and descriptions dynamically, pulling from your landing page content and existing ad copy. The AI rewrites your messaging in real time to match user intent.
Final URL expansion routes users to whichever landing page Google's model predicts will convert best, regardless of the URL you specified in your ad group. This can improve conversion rates when it works. When it doesn't, it sends high-intent traffic to pages that weren't designed for that query.
The pitch is efficiency at scale. The reality is that you're trading control for reach, and the economics of that trade depend entirely on your account's maturity, your conversion tracking hygiene, and your tolerance for variance.
The Performance Claims, Stress-Tested
Google's headline numbers come from campaigns that were "mostly using exact and phrase match keywords," where the reported lift reaches 27%. For campaigns already running broad match, the lift drops to 14%. Case studies from Think with Google cite ClickUp seeing a 20% lift in incremental conversions and 22% lower CPA after scaling AI Max across 400+ campaigns.
Those numbers are real. They're also cherry-picked.
A comparative test shared on LinkedIn by a practitioner running AI Max for four months showed a different picture: exact match delivered $52.69 per conversion, phrase match hit $43.97, and AI Max came in at $100.37. That's a 90% higher cost per conversion than phrase match, not a 14% improvement.
The pattern here isn't that AI Max is broken. It's that AI Max performs well in specific conditions and poorly in others, and Google's marketing materials don't help you distinguish which bucket your account falls into.
When AI Max Works (and When It Burns Budget)
Based on the available data and operator reports, AI Max tends to perform well when:
Your account has strong conversion tracking with sufficient volume. Google's AI needs signal density to learn. Agency testing suggests a minimum of $750/day in spend for consistent performance, though Google's official floor is $50/day.
Your campaigns are heavily constrained by exact and phrase match. If you've been running tight keyword lists and leaving reach on the table, AI Max can surface queries you'd never have thought to bid on. The 27% lift figure applies here.
Your landing pages are well-structured and conversion-optimized. Final URL expansion only helps if Google has good pages to send traffic to. If your site has thin content or inconsistent conversion paths, the AI will make bad choices.
AI Max tends to underperform when:
Your account already uses broad match extensively. The incremental reach is smaller, and you're more likely to see budget shift to lower-quality queries without a corresponding lift in conversions.
Your conversion actions are poorly defined or have long lag times. B2B accounts with 60-day sales cycles and offline conversions are particularly vulnerable. The AI optimizes for what it can see, and if your CRM data isn't flowing back into Google Ads, the model is flying blind.
Your brand or product requires tight messaging control. Text customization can generate headlines that are technically accurate but tonally wrong. If your legal or compliance team reviews ad copy, AI-generated variations will create friction.
The B2B-Specific Risks
For B2B marketers, AI Max introduces a few risks that don't show up in Google's case studies (which skew toward e-commerce and consumer brands).
First, lead quality dilution. AI Max optimizes for conversion volume at your target CPA, but it doesn't know whether those conversions become pipeline. If your form fills increase by 20% but your SQL rate drops by 30%, you've made your dashboard look better while making your revenue worse. This is a measurement problem, not an AI Max problem, but AI Max amplifies it.
Second, brand association drift. Google's documentation notes that AI Max includes brand controls at the campaign and ad group level, letting you specify which brands your ads should (or shouldn't) appear alongside. Use them. The default behavior is permissive.
Third, attribution complexity. AI Max can serve ads on queries that don't match your keyword list, using landing pages you didn't specify, with headlines you didn't write. If you're running incrementality tests or trying to map spend to pipeline, this opacity makes your job harder.
A Pilot Framework
If you're considering AI Max, don't enable it account-wide. Run a controlled test with clear success criteria.
Week 1-2: Select one campaign with stable performance history, strong conversion volume (50+ conversions/month minimum), and well-defined conversion actions. Enable AI Max with all three features active. Set up a holdout: either a duplicate campaign without AI Max or a geographic split.
Week 3-4: Monitor search term reports daily. AI Max now shows headlines and URLs in the search terms report, per Google's documentation, so you can see which queries triggered which creative combinations. Add negative keywords aggressively for irrelevant matches.
Week 5-8: Evaluate against your holdout. The metrics that matter: cost per conversion, conversion rate, and (if you have the data) downstream lead quality. A 14% lift in conversions means nothing if your sales team is drowning in unqualified leads.
Decision point: If AI Max outperforms your control on cost per qualified opportunity (not just cost per conversion), scale it. If it doesn't, turn it off and revisit in six months when Google has more training data.
The CFO Conversation
When your CFO asks whether AI Max is worth enabling, the honest answer is: it depends on your account's readiness, and the only way to know is to test it with a proper holdout.
The 14-27% lift is real for some advertisers. The 90% cost increase is also real for others. The difference isn't luck; it's account structure, conversion tracking maturity, and willingness to actively manage an AI system that requires human oversight.
AI Max is not a set-it-and-forget-it feature. It's a leverage tool that amplifies whatever signal you feed it. If your signal is clean, you'll likely see gains. If your signal is noisy, you'll scale noise.
Model it. Test it. Then decide.
