Viral claim needs an important distinction
A claim circulated online during 2026 that Google had automatically opted all users into allowing artificial intelligence to scan private Gmail messages and chats. The privacy concern behind the claim is understandable, but the description is too broad. Google told Snopes that it had not recently changed people’s settings and said it was misleading to describe the company as training the underlying Gemini generative-AI model on users’ emails.
AI processing in Gmail is not entirely new
Google services have processed account content for years when users enable features that require such processing. Gmail’s long-running “smart features” include Smart Compose, Smart Reply, automatic inbox categorization and summary cards. Tutorials explaining how to disable these features existed as early as 2020, before the current wave of generative AI. Google later expanded generative-AI capabilities across Workspace, allowing Gemini features to summarize, draft and find information when the relevant settings and account features permit it.
Google separates the controls into three settings
Google’s current help documentation distinguishes three controls. “Smart features in Gmail, Chat, and Meet” governs use of content and activity inside those services for experiences such as Smart Compose, Smart Reply, automatic categories and summary cards. “Smart features in Google Workspace” governs use of Workspace content across Workspace products, including personalized search and Gemini features that summarize content, create drafts and find key information. “Smart features in other Google products” can allow Workspace content and activity to personalize eligible experiences outside Workspace, including some features in Maps, Wallet, Gemini and Search.
Users can turn the features off, but functionality changes
On a computer, users can open Gmail, go to Settings, choose “See all settings,” and review the smart-feature controls under the General settings. Broader Workspace controls are available through “Manage Workspace smart feature settings.” Equivalent controls are available in Gmail on Android, iPhone and iPad, although some broader settings may require a web browser. Changes to the Gmail, Chat and Meet smart-feature setting apply across devices and apps where the same account is signed in. Turning settings off also disables the experiences that depend on them, including some automatic categorization, Smart Compose, Smart Reply, summary cards and cross-product personalization.
Processing content is different from training the underlying model
Google’s Workspace privacy material draws a distinction between using relevant content to provide a feature requested by a user and using that content to train the underlying generative-AI model. Google says Workspace content is not used without permission to train or improve the underlying generative-AI and large-language models that power Gemini and other systems outside Workspace. When Gemini operates inside Workspace, however, relevant content can be processed to answer a user’s request. Google also says that when smart features are enabled, Workspace content and activity may be used to improve those smart features.
Deep Research is a separate connection
The Snopes report also points to Gemini Deep Research. Google expanded Deep Research in November 2025 so users could connect sources such as Gmail, Docs, Drive and Chat and combine private Workspace information with public web research for complex tasks. That connection is separate from the long-standing Gmail smart-feature controls: users choose whether to connect Workspace sources to Deep Research. For people concerned about Gmail or Chat privacy, the practical step is therefore to review each smart-feature setting and any Gemini connections they have enabled, because turning off one control does not necessarily disable every AI-related function across a Google account.
