Google ends its five-year Gmail and Photos ad pledge, and advertisers can't see the effect
A quiet October 16 policy change lets Google use Gmail, Photos and Drive content to shape some ads in Search and AI Mode, and there is still no report that shows advertisers when it happens.

Google has rewritten the ad-privacy pledge it made five years ago, and the contents of a user's Gmail inbox and Google Photos library can now shape the ads Search shows them, if that person has opted into a feature called Personal Intelligence. The company updated its privacy policy on October 16, replacing a flat promise not to personalize ads on Drive, Gmail or Photos content, first added in July 2021, with a conditional exception for anyone who connects those apps to other Google services.
What actually changed
Personal Intelligence lets Gemini and AI Mode pull context from a user's email, photos and files: a flight confirmation sitting in Gmail, for instance, can prompt a travel suggestion elsewhere in Search. Google launched the feature in January as a Labs experiment for AI Pro and AI Ultra subscribers in the US, then widened it to roughly 200 countries at I/O in May, excluding the EU, UK, Switzerland and Nigeria. Through March, Google said the experience would stay ad-free. That changed around October 16, when the company confirmed it had begun testing ads inside AI Mode for people who have Personal Intelligence turned on.
Google's own help documentation draws a careful line around how that works. A handful of protections carry over unchanged: advertisers still can't see inbox or photo contents directly, and Google still won't personalize ads on race, religion, sexual orientation or health. The new policy also adds a third basis for ad personalization alongside the old standbys of interests and activity: "personalized content from Google services," meaning summaries and inferences pulled from whatever a user has connected.
"That same info may be used to show you ads," Google's help documentation says, after noting ads won't be based directly on Gmail or Photos content.
The mechanics matter for how long this follows a user around. Google says its Search services don't train models directly on raw inbox contents, but they do train on the summaries generated from that content, and a subset of what Personal Intelligence processes in AI Mode and Ask Maps gets reviewed by human staff. Disconnecting Gmail or Photos later doesn't delete data Google has already collected and summarized -- it only stops new data from flowing in. None of this is exotic by AI-platform standards, but it's a meaningfully different posture from the flat no-personalization promise Google has pointed advertisers and users to since 2021.
Who it touches
Google has named the campaign types eligible for this ad test: Performance Max, Shopping and broad-match Search, including AI Max for Search. That covers most of the demand-generation budget a software company runs through Google, not just retail and travel spend. If a prospect has connected their Gmail or Photos to Search, the signals that put an ad in front of them may now include inferences drawn from their personal accounts, alongside whatever keyword or audience targeting a marketer actually set.
What hasn't shown up anywhere is a way to measure any of it. Google has published no figures on how many users have connected their apps, and neither the revised policy nor its help pages describe a new targeting option, audience segment or report tied to this signal. AI Max search-term, asset and landing-page reports don't flag whether a connected-app inference contributed to a given impression or conversion, so an advertiser has no way to isolate the effect, positive or negative, on their own numbers. Retail and travel marketers are the obvious early test group, since Google's own hotel-booking example leans that way, but the eligible campaign types reach well past those categories. A B2B software company running broad-match Search or AI Max alongside its demand-gen stack is just as exposed to the signal, and just as blind to it, as a consumer retailer.
Where the opt-out actually lives
The control surface is easy to miss, too. Turning off personalized ads in My Ad Center, the interface Google said had more than 60 million daily users as of January 2024, does not stop Gmail or Photos data from feeding Personal Intelligence. The only way a user fully removes that data from AI features is to disconnect the apps under Search's own personalization settings, a separate control most people have never opened. Google is also folding its older Web & App Activity and Search Personalization settings into two new controls, Search Services History and Personalized Recommendations, over the coming months, which will move the relevant switches again.
- No new targeting option, audience segment or report comes with this signal
- Sensitive categories (health, religion, race, sexual orientation) stay off-limits
- Disconnecting apps under Search settings is the only full opt-out; My Ad Center alone isn't enough
Don't expect new targeting upside from this test; there's no reporting that lets you use it on purpose yet. Treat it instead as a disclosure risk. If a prospect asks whether your Search or Performance Max ads are drawing on their Gmail, the honest answer today is that neither you nor Google's own reports can confirm or rule it out. Flag that gap to whoever owns privacy and trust messaging before a customer raises it first, and check back for a dedicated reporting breakout — its absence right now is the detail worth re-checking next quarter.
PositioningAnthropic's IPO filing shows sales and marketing spend up 26x in six months
Sienna Mcpherson · 4 min read
Positioning1Password launches a trust layer for AI agents, and its own growth numbers back the bet
Sienna Mcpherson · 4 min read
Distribution