Google ties Gmail and Photos data to ads via Personal Intelligence
Google's advertising section has said the same thing for five years: personalized ads are not built from the contents of your Drive, Gmail or Photos. As of October 16, that sentence comes with an exception — and the exception is the AI feature the company spent this year persuading people to switch on.
One clause, four edits
The old bullet read: "We don't show you personalized ads based on your content from Drive, Gmail, or Photos." The new one keeps every word and then continues — "...unless you choose to make data from these services available to other Google services that support personalized ads." Google publishes a redline against the previous version in its privacy policy archive, and PPC Land's line-by-line comparison of that diff against 54 earlier versions, running back to 1999, found the promise had stood in essentially identical wording since July 2021, per PPC Land. The change was first spotted by Android Authority and picked up by The Verge, Ars Technica and 9to5Google on October 23, a week after it took effect.
That bullet is not the only edit, though it is the only one being quoted. Four passages in the "provide personalized services, including content and ads" section changed. One adds a new sentence explaining that Search or Gemini Apps may read your other Google services to give more personal answers, and that "Search may process some of the data that you allow Search to access to personalize content, including ads." Another widens the general basis for ad targeting from "your interests and activity across Google services" to "your interests, activity, or personalized content from Google services."
The mechanism is not mysterious. Personal Intelligence, which began rolling out in beta in January, is the opt-in that lets Google's AI products read Gmail, Drive, Photos and Search history to answer questions with your own context. It is genuinely useful — ask about dinner near a hotel you booked and Search can find the confirmation email. It is also the pipe through which inbox contents reach a surface that carries advertising.
The edit that isn't in the headlines
The third basis is the one worth reading twice: personalized content now stands alongside interests and activity as a reason an ad can be aimed at you. Interests and activity are behavioural — what you clicked, what you searched. Personalized content is a different class of signal. It is the model's summary of you, produced from documents you never intended to be readable by an ad system, and it can say things your click history cannot: that you are moving flats, that a family member is ill, that you are negotiating a salary.
Google also added a note to the policy's Additional Context section defining "personalize content." The note names Personal Intelligence directly, and gives the hotel-booking example. It does not mention ads. The sentence it hangs off does. Read separately, neither passage tells you what the combination means; the scope of the change lives in the gap between them, which is a strange place to put a definition that governs advertising.
Why the timing is a coincidence only on paper
Google has an ads problem specific to AI: the answer surface it is betting on is the one that removes the click. AI Mode answers the question, so there are fewer result pages to sell. Ars Technica notes that AI search tools may show sponsored content while the Gemini app itself carries no ads — which makes ad performance inside AI Mode the thing Google has to prove, and targeting the lever it has. An untargeted sponsored slot inside a conversational answer is worth very little to an advertiser; one informed by the user's own hotel booking is worth a great deal.
Compare the approach OpenAI took, which we went through in ChatGPT ads: 100,000 advertisers and a $1B run rate in 200 days: an ads service isolated from the answer model, with an orchestrator that reads the conversation and nominates candidates into a separate auction. That system's input is the chat. Google's input can be years of correspondence. The engineering problem is smaller; the permission problem is much larger.
What Google says, and the word carrying the weight
Google's statement to Ars Technica is that connecting apps is entirely opt-in, that "ads will continue to be relevant to things like your query, the context of the response and your interests," and that "ads will not be based directly on your Gmail inbox or Photos library," Ars Technica reports. A spokesperson also told Ars the clarification does not change how data is handled with respect to advertising partners.
Both statements can be true and still describe the thing people are worried about. "Directly" is the operative word: if Search reads the booking email, summarises it, and then serves a sponsored restaurant suggestion near that hotel, no advertiser saw your inbox and no ad was drawn straight from it. The chain is real anyway, and the policy is now written to permit it. The unchanged protections are genuinely unchanged — no ad targeting on sensitive categories such as race, religion, sexual orientation or health, and no passing your name or email address to advertisers — but they constrain what advertisers see, not what Google's models may read on the way to choosing an ad.
The unresolved question is consent, and it is unresolved rather than settled. Personal Intelligence is opt-in, but the opt-in arrives through prompts and pop-ups, several taps deep, at the moment a user wants an answer. Whether a toggle presented that way meets the GDPR's bar for specific, informed consent to ad personalisation is exactly the argument European regulators have been having with Meta over "pay or consent" — and Google has now written an ad-personalisation purpose into a feature whose consent flow was designed for convenience, not advertising. No regulator has acted on this change. That is what makes it worth watching rather than concluding.
Check your own account
Personal Intelligence's connected apps are the switch. In your Google Account, go to Data & privacy, then Personalization settings, then Manage Personalized Recommendations in Search services, and look for Connected Content apps; disconnecting Google Workspace and Google Photos there stops Search reading them. Turning off Personalized Recommendations entirely is the broader option. One caveat from Android Authority's walkthrough: disconnecting an app does not necessarily remove what Google already stored — data in Search Services History can persist after you disconnect.
Worth pairing with the other half of the same trade: OpenAI's ad pixel tracks ChatGPT users across 1,000 websites. The two companies are converging on the same bargain from opposite ends — OpenAI is building a profile from the outside in, Google from the inside out.
What to watch
Whether ads appear in the Gemini app itself, which is the point at which "the assistant has no ads" stops being the reassurance. Whether Google publishes any number for how many accounts have connected Workspace to Personal Intelligence, which it has not. And whether the phrase "personalized content" gets defined more tightly anywhere a reader would find it — because right now the definition sits in a linked note that talks about hotels and not about advertising.
If you had Personal Intelligence on since January, did you know ads could draw on it — and would you turn it off now? Write to us via our contact page.
