Google Adds an Opt-In Exception to Its Promise Not to Personalize Ads From Gmail, Drive and Photos, and Its Help Page Says Ads May Use “Summaries, Excerpts and Inferences”

Google’s Privacy Policy, effective October 16, now says it does not show personalized ads based on your content from Drive, Gmail or Google Photos “unless you choose to make data from these services available to other Google services that support personalized ads.” Before the update the sentence ended after the word Photos. That unconditional form is in each of the five archived policy versions we checked, from January 15, 2024 to October 1, 2026.

Two related changes landed around the same time. A Google Search Help page on connecting Gmail and Photos to Search now says ads “will not be based directly on your Gmail inbox or your Google Photos library,” but also that Search may reference “summaries, excerpts and inferences” from them to personalize your experience and that “that same info may be used to show you ads.” And on Friday, October 16, Google told Search Engine Land that it is starting to test ads in AI Mode for people who connect their apps through Personal Intelligence, according to Search Engine Roundtable. Ars Technica reported the policy change on October 23.

We worked from Google’s own policy archive, Help pages, launch posts and Internet Archive captures, so most quotations below come from Google’s pages, not from coverage. Google’s statement on the policy reaches us through Ars, and its statements to Search Engine Land through Roundtable, because Search Engine Land answered our fetcher with a challenge page.

What changed in the policy

Google’s policy archive has a comparison page that marks insertions and deletions. Between the October 1 and October 16 versions it marks nine changed passages. Besides the effective date and three small edits (the word Google added before Photos in one sentence, and two Learn more links dropped), these are the changes, with the last row combining a new list entry and its explainer:

Two neighboring promises did not change. Google still says it does not show personalized ads based on sensitive categories “such as race, religion, sexual orientation, or health,” and does not share information that personally identifies you with advertisers “such as your name or email, unless you ask us to.” The ads paragraph does gain an input: ads may now be based on “personalized content from” Google services, where the old text named only interests and activity.

Where “directly” comes from

The qualifier Google leans on is not in the policy’s new text, whose only uses of “directly” are in unrelated sentences about contacting you and resolving complaints. It is on the Search Help page Connect your Google content apps to get a Search services experience that’s tailored to you, which is the page the policy’s new Learn more about Personal Intelligence link opens. That page now answers two questions about ads:

“If I connect my content apps, will my Gmail inbox or Google Photos library be used to serve ads? Ads will not be based directly on your Gmail inbox or your Google Photos library. If you connect your content apps to Search Services, Search may reference summaries, excerpts and inferences from these sources to personalize your experience and do things for you. That same info may be used to show you ads, so they are even more helpful for you. As always, Google doesn’t show you personalized ads based on sensitive categories, such as race, religion, sexual orientation, or health. You get to choose if you want to connect your content apps to Search Services. You can adjust your settings, including disconnecting content apps from Search services, at any time.”

“If I connect my content apps, will you share my Gmail inbox or Google Photos library with advertisers? No. If you connect your content apps to Search Services, advertisers will not have access to your Gmail inbox or Google Photos.”

An Internet Archive capture from October 11 at 20:41 UTC of the same page mentions neither ads nor advertisers; a capture from October 16 at 19:36 UTC has the answer. We checked only those two captures, so the edit can be dated no more finely than that window.

Google’s statement as Ars prints it is in the same terms: “Connecting your apps is entirely opt-in. Ads will continue to be relevant to things like your query, the context of the response and your interests. Ads will not be based directly on your Gmail inbox or Photos library, and as always, sensitive information will not be used to tailor ads.” In Ars’s words, a Google spokesperson said the clarification “does not indicate a change in how it handles your data with respect to its advertising partners.”

The word is not new in Google’s Personal Intelligence language. The January 14 launch post says Gemini “doesn’t train directly on your Gmail inbox or Google Photos library” and that Google trains “on limited info, like specific prompts in Gemini and the model’s responses.” The January 22 AI Mode post and the March 17 expansion post repeat the formula.

A timeline of the promise

Who is affected, according to Google’s pages

• It is opt-in. Google’s Gemini launch post says “Connecting your apps is off by default,” and its Search post calls connecting Gmail and Photos “strictly opt-in.” Gemini Apps Help describes a card with Connect all, Choose apps and Don’t connect options.
• It is for personal accounts. Google’s posts say the connected experiences are for “personal Google accounts and not for Workspace business, enterprise or education users,” and Gemini Apps Help says the feature “isn’t available when you sign in to a work, school, or supervised Google Account.” The policy itself sends members of organizations that use Google Workspace or Google Cloud Platform to the Google Cloud Privacy Notice.
• One connection covers many products. The Search Help page says: “Connecting your Google content apps means that all Search services will be permitted to access Workspace (currently Gmail and Google Calendar) and Google Photos. Search services include Google Search (incl. AI Mode, Discover), Maps (incl. Ask Maps), Shopping, Flights, Hotel, Translate, and News.” The only ads test reported so far is in AI Mode.
• It is offered in a limited set of countries. The supported-country list on the same page includes the United States, Canada, India, Japan, Australia and Brazil. We did not find the United Kingdom, Switzerland, Germany, France, Italy, Spain, Ireland, Poland or Sweden on it.

What the pages leave open

• Which ads. The policy says “other Google services that support personalized ads” and gives no list. The Help page speaks of Search services. Ars says the Gemini app does not currently have ads, and the one test reported is in AI Mode.
• What an inference is. The Help page names “summaries, excerpts and inferences” without defining them or giving a retention period. The boundary it states is the sensitive-category exclusion.
• The ad controls. The policy says you “can control what information we use to show you ads by visiting your ad settings in My Ad Center.” Neither the Search Help page nor Gemini Apps Help mentions My Ad Center, so we cannot say whether turning off ad personalization stops connected-app inferences from reaching ads. We did not test it.
• The test. None of the sources we could read says in which countries the AI Mode ad test runs, how many people see it or which ad formats it uses.
• The change clause. The policy promises “We will not reduce your rights under this Privacy Policy without your explicit consent” and, if changes are significant, “a more prominent notice (including, for certain services, email notification of Privacy Policy changes).” Google’s position, as Ars relays it, is that connecting apps is opt-in. Whether people who connected apps before October 16 consented to ads, and whether Google sent email notices for this update, are not visible from public pages.

To see what is connected in Gemini, Google’s help page says to open gemini.google.com, then Settings & help, Personal Intelligence and Connected Apps. For Search, the January 22 post says to open Search, tap your profile, then Search personalization and Connected Content Apps; the Search Help page warns that its settings “are being updated so may not match your current experience.” For an organization, Google’s pages put the choice with each person’s personal account, and we found no administrator control described in them.

Our reading

The promise changed shape. The old line was a rule about a source: nothing from Drive, Gmail or Photos feeds personalized ads. The new line is a rule about a path: nothing does unless you connect those services to others that show personalized ads. That is a narrower promise, and the policy says so in plain words.

“Directly” does the work, and it sits outside the policy. The policy’s own example already says Search may process data you allow it to access “to personalize content, including ads.” What limits that is a Help page answer saying ads are not based directly on your mail, while summaries, excerpts and inferences may feed them. Nothing we read defines directly. The same word has qualified Google’s training promise since January, so separating raw content from what is derived from it looks like a standing way of writing these assurances, not a one-off.

Both headlines can be defended from the documents. Ars’s “your Gmail might be used for advertising” fairly summarizes a policy that now has an exception, and Google’s “will not be based directly on your Gmail inbox” fairly summarizes its Help page. The disagreement is over what directly covers, and the policy leaves that to a Help page we could date only to a window between two archive captures. Google flagged this direction in March, when it said it anticipated ads would operate similarly for people who connect apps; the October pages are where that became text. For another AI assistant’s path to ads, see our earlier post on ChatGPT Ads.

Sources and method

The policy text and comparison are from Google’s Privacy Policy and its archive, fetched between 10:05 and 10:20 UTC on October 25, together with the archived versions of January 15, 2024, July 1, 2025, December 11, 2025, April 2, 2026 and October 1, 2026. We read the English-language (U.S.) version and did not compare regional versions. The Help pages are Google’s Search Help and Gemini Apps Help. Search Engine Roundtable’s article is dated October 19 and Ars Technica’s October 23 (Ryan Whitwam); Ars credits Android Authority with spotting the change, and we did not retrieve that report. Search Engine Land’s pages, including the March article that Roundtable quotes, returned a challenge page to our fetcher, so those quotations are Roundtable’s.

We did not sign in to any Google account, connect any app, see an ad or see the in-product screens where people connect their apps; every statement about product behavior comes from Google’s documentation or the outlets cited. The country check is a text search of the list on the Search Help page. We used the same archive-comparison method on OpenAI’s Usage Policies in an earlier post.