A future iteration of Google’s Gemini AI model, known as “Personalization,” is set to introduce a feature that allows the model to access users’ Google Search history to provide more tailored and relevant responses. This development, uncovered by software analyst AssembleDebug, was found within the latest beta release of Android 16.
This new Gemini model will differ from its predecessors by requiring explicit user permission before accessing search history data. Users will also need to ensure their “Web & App Activity” setting is enabled to facilitate this connection.
The primary function of this feature is to enable Gemini to respond to queries that rely on past search activity. For example, users could ask questions like, “When did I last search for [specific item]?” or request a summary of their search history. Additionally, the model may offer functionality such as recommending search history entries for deletion or identifying potential personality traits based on search patterns.

However, this integration raises concerns regarding potential privacy and security implications. Access to detailed search history data could lead to misinterpretations or misuse of information, particularly in scenarios involving law enforcement or other third-party access. Google has addressed some of these concerns by stating that the “Personalization” model is the only Gemini variant that will connect to search history and that users can disconnect this feature at any time.
Furthermore, Google has indicated that conversations with the “Personalization” model will not be used to improve Gemini’s algorithms and will not be stored beyond the user’s chat history. These chats will be automatically deleted from Google Apps Activity after 60 days. Search history data will also remain isolated from other Gemini models.
According to Google, “This version of Gemini uses your Google Search history to give you personalized help. In order to use it you’ll need to first connect Gemini to your search history. You can always disconnect it later.”
The discovery of this feature in the Google app’s beta version suggests that Google plans to launch the “Personalization” model in the near future. Tests have demonstrated the model’s ability to retrieve search data dating back to the beginning of the current year.