Your AI assistant just learned to remember your coffee preferences and your cat’s name, but it also gained the ability to forget everything on command. Google Gemini’s latest update introduces two seemingly contradictory features that reshape how you’ll interact with AI going forward.
Personal Context Turns Chat History Into Recommendations
Your past conversations now fuel smarter, tailored responses across all future interactions.
Gemini’s new “Personal Context” feature automatically analyzes your chat history to deliver personalized recommendations and responses. Ask for book suggestions, and it’ll remember you mentioned loving dystopian fiction three months ago. Request restaurant recommendations, and it’ll factor in your previously stated dietary restrictions.
The feature comes enabled by default on supported accounts, though you can toggle it off in settings if you prefer your AI assistant to maintain goldfish-level memory. This replaces Google’s previous “Saved Info” system, making personalization automatic rather than user-driven.
Temporary Chats Offer True Privacy Mode
New incognito-style conversations disappear within 72 hours and never influence future responses.
For sensitive discussions or one-off queries, Temporary Chats provide genuine privacy protection. These conversations won’t appear in your chat history, won’t train Google’s AI models, and won’t influence future personalization. Think of it as incognito mode for AI conversations.
The catch? Google retains this data for up to 72 hours before deletion, allowing you to review or continue discussions. You’ll find the Temporary Chat button right next to “New chat” in the app.
Privacy Controls Get More Granular
New opt-out options give you precise control over how your conversations train AI models.
Google renamed “Gemini Apps Activity” to “Keep Activity” and updated its data usage policies. Starting September 2, chat samples may train AI models unless you explicitly disable “Keep Activity” or exclusively use Temporary Chats.
This granular control addresses growing concerns about conversational AI data practices. You can now choose exactly what level of personalization and data sharing feels comfortable, from full context awareness to complete conversation amnesia.
The Privacy-Convenience Balance
These features acknowledge that different conversations require different privacy levels.
The rollout starts with Gemini 2.5 Pro users in select countries (notably excluding the EU and UK initially) and requires users to be 18 or older. This dual approach—smarter personalization alongside robust privacy controls—reflects the reality that you might want Gemini to remember your work preferences while forgetting your medical questions.
Your choice of chat mode now determines exactly how much your AI assistant learns from each conversation.