Connecting Gmail or Google Photos to Google's Personal Intelligence can do more than tailor an AI answer. Google's current documentation allows information derived from those connections to help select ads. If you enabled the feature for convenience, the useful next step is to review both the permission you granted and the information already saved.
The Google Privacy Policy effective October 16, 2026 describes a conditional exception for content made available to other Google services that support personalized advertising. The Verge covered the issue on October 23. Those dates differ: the later article did not itself mark the policy's start.
The ad test also has its own timeline. Search Engine Land reported ad-free connected-app sessions in March and Google's confirmation of an ads test on October 16. The latter report did not establish a general rollout date.
The permission concerns information derived from your content
Google's Search help says ads are not based directly on the Gmail inbox or Photos library. It also says summaries, excerpts and inferences used to personalize Search may inform ads. The practical distinction is between giving an advertiser access to your files and letting Google use derived context when selecting an advertisement.
Google says advertisers do not gain access to the connected inbox or photo library and that sensitive categories such as health, religion and sexual orientation are excluded from personalized ads. These are the company's stated limits, not findings from an independent TechKili audit.
The connection is optional. Nor should a user assume that permission is confined to one AI Mode answer: the help page describes access across Search services, including Maps, Shopping and other listed services. Available connections vary by region and account conditions.
Review three controls for three different purposes
A useful review separates the following choices:
Connected Content Apps determines which supported content services you connect to Search.
My Ad Center manages advertising personalization.
Search Services History concerns activity already retained; Google links to its activity controls for managing and deleting it.
The same help document explicitly says disconnecting an app does not delete data already stored in Search Services History. That retained data can still support product improvement, including model training. Disconnecting a source and managing its previously saved output are separate actions.
Google warns that its settings are being updated, so labels and navigation may differ from the documentation. Follow the current controls for the signed-in account instead of assuming an older menu path remains exact.
Make the connection decision with the full terms in view
The help page also describes training and human review of some connected-app-derived interactions. It notes that a short or particularly relevant email or file may itself serve as the summary. These existing documented conditions deserve attention alongside the advertising change.
Someone who does not want that use of their content can leave apps disconnected and still use AI Mode. That choice should not be interpreted as an ad-free Search promise. The decision is whether the benefit of personalized answers justifies this additional use of personal context, followed by a separate review of retained history.
Sources
Related Articles
Apollo's AI payoff claim hinges on how productivity is measured
Torsten Slok's October 24 note questions broad AI returns. Its index comparisons and productivity charts are useful context, but they do not measure a particular company's deployment.
Etsy bars AI-written functional guides and asks listings to show the real result
Etsy's October policy update separates permitted AI-assisted art from functional instructions, while tightening the evidence sellers must show in product images.
vLLM TorchTPU reuses PyTorch models, with validation tied to each TPU
The new backend keeps vLLM’s model definitions and serving interface. Hardware coverage, checkpoint validation and compilation still need separate evaluation.
RTX Spark brings CUDA PyTorch to Windows Arm64, with dependency gaps
Nvidia documents native Windows Arm64 builds as RTX Spark laptops arrive. Developers still need to match Python wheels, extensions and drivers to the complete workload.