Asking an AI assistant to find a hotel reservation in your inbox is not quite the same as agreeing to let information from that reservation influence advertising.

The same personal context may technically support both activities.

But the purposes are different, and the distinction matters to users.

Google's latest privacy policy revision brings that issue into focus.

The updated terms became effective on October 16, 2026. Technology publications including Android Authority, Ars Technica, The Verge and 9to5Google examined the change on October 23.

The revision introduces a condition to Google's established promise not to personalize advertisements using content from Gmail, Drive or Google Photos.

The company now makes an exception when users choose to share data from those services with other Google products that support personalized advertising.

That may sound narrow, but it becomes important when considered alongside Google's growing ecosystem of AI-powered personal assistance.

What changed in the October 16 privacy policy

Google's previous policy contained a straightforward protection.

Content from Gmail, Drive and Google Photos was excluded from personalized advertising.

The current policy retains that general restriction while adding a conditional exception.

If users authorize another Google service to access information from those applications, the information may be processed for personalization under the receiving service's settings and policies.

Google explicitly identifies Search and Gemini Apps as examples of services that may receive such information.

Its policy also explains that Search may use certain authorized data to personalize content, including advertising.

The legal and practical scope of the promise has therefore changed.

Previously, the advertising exclusion did not contain this cross-service exception.

Now, its applicability depends partly on which connections users have enabled.

That does not establish that every Gmail account has automatically become available for advertising analysis.

The relevant provision describes conditional access linked to user choices.

Personal Intelligence connects private context to AI services

The change is best understood through Personal Intelligence.

Google introduced the feature in Gemini in early 2026 and subsequently expanded it into AI Mode in Search.

The idea is to make AI assistance more useful by allowing connected applications to provide personal context.

Instead of manually copying information from a message, a user can authorize the assistant to retrieve relevant details.

The same approach can incorporate photographs, documents and information from other Google services.

For example, an assistant might use a travel confirmation to understand when someone plans to visit a destination.

It could then provide recommendations without requiring the person to restate every detail.

This addresses a fundamental limitation of generic AI assistants.

Without personal context, they can answer general questions but often struggle to provide genuinely relevant help.

Connecting applications reduces that friction.

It also allows information originally stored for one purpose to influence interactions with another service.

Personal context can indirectly shape advertising

The concern is not necessarily that an advertising system directly scans every email for keywords.

The revised policy creates room for a more indirect process.

Consider a user who connects Gmail to AI Mode in Search and requests recommendations for an upcoming trip.

The assistant may use authorized reservation information to determine a destination and travel dates.

Its answer can then reflect personal details that were not included in the original search query.

If advertising is displayed within the same experience, the resulting context may contribute to determining which ads are relevant.

This is an illustration of the policy's potential implications, not evidence that a particular advertising campaign has used private reservations.

Ars Technica's analysis focuses on precisely this distinction.

The question is no longer limited to whether Gmail itself provides data directly to an advertising engine.

It also concerns whether insights derived from Gmail can affect advertising in another Google service.

Google says ads are not based directly on private inboxes

Google provided a clarification to Ars Technica following questions about the new language.

The company says connecting applications remains entirely optional.

It also says ads continue to reflect signals such as the user's query, the context of a response and their interests.

According to Google, advertisements are not based directly on the contents of an individual's Gmail inbox or Google Photos library.

Google further states that the clarification does not indicate a change in how information is handled with advertising partners.

Advertisers do not receive direct access to a person's messages or photographs simply because of this policy revision.

The policy also maintains restrictions on sharing directly identifying information, such as a name or email address, without permission.

These assurances are meaningful.

However, they do not entirely answer questions about indirect personalization through connected AI services.

A system can avoid providing direct access to private messages while still incorporating contextual signals derived from them into another service.

Understanding that distinction is essential when evaluating the impact of the new terms.

Sensitive advertising categories remain restricted

Google's updated policy continues to exclude certain sensitive categories from personalized advertising.

These include health, religion, race and sexual orientation.

The new exception for connected services does not remove that stated restriction.

Nevertheless, not every potentially revealing piece of information falls into a formally sensitive category.

Travel plans, purchases and everyday preferences can provide detailed insights into someone's life.

When combined across services, seemingly ordinary information may become substantially more revealing.

This makes the scope of permitted data processing important even when sensitive-category protections remain in place.

The revised policy establishes conditional access and personalization but does not publicly describe every internal transformation between source data, generated responses and advertising signals.

It would therefore be inappropriate to claim detailed knowledge of Google's underlying implementation.

Consent exists, but does users' understanding keep pace?

Google emphasizes that Personal Intelligence connections require user permission.

People can decide whether to enable access to particular applications.

This is an important protection against indiscriminate sharing across the company's entire ecosystem.

Yet technical consent does not automatically guarantee informed understanding.

Someone may connect Gmail to Search because they want an assistant to retrieve information more conveniently.

They may not expect that decision to influence advertising personalization in the same service.

The updated policy explicitly connects those possibilities.

That raises a practical question about interface design.

Does an authorization described as improving assistance clearly communicate its potential advertising implications?

Google states that data use remains subject to the policies and settings of the service receiving the information.

Users may consequently need to understand several layers of controls distributed across different products.

How to review connected applications

Users who want to limit cross-service sharing can inspect their Google account settings.

Gemini provides Personal Intelligence controls for reviewing and changing connected applications.

Google Account settings also include controls for personalized recommendations in Search services.

Ars Technica describes a route through Data and privacy, personalization settings, personalized recommendations in Search services and connected content applications.

The exact labels may vary according to language, region and interface changes.

Users can review connections involving Google Workspace and Google Photos and disable access they do not need.

This does not necessarily require abandoning the underlying Gmail or Photos services.

However, connected-application controls are not identical to advertising preferences.

Disconnecting an application limits a particular route for sharing data.

My Ad Center offers separate controls related to personalized advertising.

Users should not assume that changing one setting automatically disables every form of personalization across Google.

Gemini and Search do not necessarily handle advertising identically

Another distinction concerns the products involved.

Personal Intelligence can operate in multiple Google services, but their advertising capabilities are not identical.

Google's comments to Ars Technica distinguish the Gemini app from AI-powered Search experiences where advertising may appear.

Allowing Gemini to use personal context does not by itself prove that an advertisement based on that context will appear inside the Gemini application.

The concern is particularly relevant to services that combine AI-generated answers with advertising.

Google Search is an obvious example.

The general privacy policy now explicitly permits certain cross-service personalization scenarios.

That does not establish that every Google product implements the same advertising behavior immediately.

AI training is a separate privacy question

The October 16 revision should not be confused with Google's policies governing AI model training.

When Personal Intelligence was introduced, Google said it did not train its models directly on entire connected Gmail inboxes or Google Photos libraries.

The company distinguishes retrieving information for an individual response from using certain interactions to improve its systems.

Those activities are subject to different settings and product policies.

Advertising personalization represents another separate purpose.

A meaningful privacy assessment should therefore ask several questions independently.

Which data can the assistant access? How long is information retained? What is it used for? Which settings control each activity?

The current policy change primarily concerns cross-service personalization that may include advertising.

It does not establish a blanket change to every Google AI training practice.

Why this exception matters for the future of AI assistants

Technology companies increasingly want assistants that understand users' circumstances rather than responding only to isolated questions.

Connecting services can make those systems substantially more useful.

It can also provide commercial advantages to platforms that manage several different categories of personal information.

Google operates communication, cloud storage, search, video and advertising services.

Connecting those environments creates significant opportunities for personalization.

The important issue is not simply the amount of available data.

It is whether information provided for one purpose can be repurposed for another commercial function.

A message stored in Gmail for communication or a photograph saved as a personal memory was not necessarily provided with advertising in mind.

Personal Intelligence makes those connections more useful, but potentially more difficult to evaluate.

The new exception makes the privacy implications of cross-service integration especially important.

A conditional protection users should understand

Google's October 16 revision does not establish that the company sells private Gmail messages or photo libraries to advertisers.

Nor does it demonstrate that every Google user has automatically enabled Personal Intelligence.

What it does establish is a meaningful qualification to a previously unqualified advertising privacy promise.

Information from Gmail, Drive and Photos can potentially contribute to personalization, including advertising, when users authorize access through compatible Google services.

Google says those connections remain optional and that sensitive information is not used to tailor ads.

The practical difficulty lies in understanding the consequences of enabling multiple interconnected features.

Privacy-conscious users can review connected applications, disable unnecessary access and examine their advertising preferences separately.

For Google, the broader challenge is demonstrating that more personally aware AI assistance can coexist with clear, understandable boundaries between helpful functionality, data processing and advertising.