Google spent months promising that users who connected Gmail, Calendar, and Photos to AI Mode would get an ad-free experience. That promise expired in October 2026. Search Engine Land reported that Google began testing ads in AI Mode specifically for users with Personal Intelligence enabled, reversing the earlier stance. For demand gen teams running Google paid media, the shift raises a measurement problem that matters more than the targeting change itself.
What Actually Changed
Personal Intelligence is Google's opt-in layer that lets AI Mode pull context from connected apps (Gmail, Photos, Calendar, Drive) to personalize responses. Until this test, those users saw no ads. Now they do. The signals shaping which ad appears have expanded: the query, the AI response context, the user's interests, plus summaries, excerpts, and inferences derived from connected apps.
Google says ads aren't based directly on someone's Gmail inbox or Photos library, and advertisers don't get access to connected-app data. But The Verge flagged that Google's updated privacy policy now includes an exception: data users choose to make available from services like Gmail and Photos can be used by other Google services that support personalized ads. The practical effect is derived-signal personalization. Google infers context from your connected apps and uses those inferences to select ads. Advertisers never see the raw data.
That's an important nuance. It also doesn't help you measure anything.
The Measurement Gap Is the Real Story
CTR data, conversion-lift numbers, published performance benchmarks for AI Mode ads in this context: none exist. Google hasn't released them. Independent studies haven't produced them either.
What we do have is ad prevalence data. An SE Ranking study of 50,032 ad-eligible commercial keywords found that ads appeared in 29.45% of AI Mode responses. For keywords with CPCs of $10 or more, that figure jumped to 53.6%. Among responses showing an ad, 71.1% displayed two advertiser offers. These numbers tell you ads are showing up. They don't tell you whether those impressions produce pipeline.
Google's own user survey stat (75% of people said AI Mode helped them make faster, more confident decisions) measures satisfaction with AI Mode itself, not ad performance. Treating presence data as proof of effectiveness is exactly the attribution shortcut that gets demand gen leaders in trouble with their CFO.
What You Can Do This Week
There's no AI Mode campaign type. Eligibility flows through Performance Max, Shopping, and Search campaigns with broad match (including AI Max for Search). If your account runs only exact and phrase match, you're likely invisible to this inventory. That's your first diagnostic: check campaign structure against eligibility.
The hypothesis worth testing: if we shift one high-intent campaign to broad match with tighter conversion signals, then qualified lead volume from Google will change (direction unknown) because the auction now includes AI Mode placements we weren't reaching. Set a two-week window. Primary metric: cost per qualified opportunity. Guardrails: CPA increase capped at 25%. Stop-loss: if CPA rises above that threshold for five consecutive days, revert.
Three operational moves that don't require waiting for Google to publish performance data:
- Clean your conversion data. If you're still passing form fills as your primary conversion, you're telling the algorithm to find people who fill forms, not people who become qualified opportunities.
- Audit paid-organic consistency. AI Mode blends answers, links, and ads on the same screen. Pull five real customer questions from sales calls, run them in AI Mode, and note where your brand shows up (or doesn't) in both answers and ads.
- Build landing pages for follow-up questions. AI Mode conversations are multi-turn. A page optimized for a three-word keyword won't satisfy someone mid-conversation about a complex purchase. Structure pages around the next question, not the first one.
The Trust and Governance Layer
For B2B teams in regulated categories, the privacy-policy update deserves a compliance review. The opt-in nature creates uneven exposure across your audience, which complicates forecasting and experiment design. You can't segment reporting by "Personal Intelligence users" versus everyone else, because Google doesn't expose that dimension to advertisers.
There's also a brand-safety question. When ads appear inside AI-generated answers, users may perceive those answers as neutral guidance rather than a mix of editorial and paid content. Vietnam.vn raised the question of how users distinguish advice from sales pitches when both live in the same response. Ad copy in AI Mode needs to be more substantive and less promotional than a standard search ad, because the context around it looks like trusted information.
Where This Leaves You
Google is moving search toward derived-signal personalization, where the system infers intent from connected-app context rather than relying solely on typed queries. Advertisers get no new levers, no new targeting options, and no performance data to evaluate the change.
That's not a reason to opt out. It is a reason to treat this as an experiment with defined guardrails rather than a budget reallocation. The teams that build clean conversion data, consistent paid-organic messaging, and landing pages designed for conversational queries will be positioned well regardless of whether Google ever publishes the lift numbers. The ones chasing a new placement without measuring its downstream impact will discover, again, that presence in a feed isn't the same thing as pipeline.
