Google Gemini is now able to access your search history.

      Google has unveiled a variety of enhancements for its Gemini assistant today. Firstly, the updated Gemini 2.0 Flash Thinking Experimental model now supports file uploads as input and has received a performance boost.

      A key update is a new opt-in feature named Personalization. In essence, when you ask a question to Gemini, it reviews your Google Search history and delivers a customized response.

      In the future, Personalization will extend beyond just Search. Google states that Gemini will also utilize other apps within its ecosystem, like Photos and YouTube, to enhance its personalized responses. This is somewhat reminiscent of Apple's postponed AI features for Siri, which even led the company to withdraw its advertisements.

      Regarding the Google Search integration, when you request recommendations for nearby cafes from the AI assistant, it will check your past searches for that information. If relevant data is found, Gemini will incorporate those details and business names into its response.

      "This will enable Gemini to provide more personalized insights, leveraging a wider understanding of your activities and preferences to deliver responses that genuinely resonate with you," notes Google in a blog entry.

      The new Personalization feature is connected to the Gemini 2.0 Flash Thinking Experimental model and will be available to both free and paid users with a Gemini Advanced subscription. The rollout starts today, initially for the web version, and will soon extend to the mobile application as well.

      Google indicates that the Personalization feature currently supports over 40 languages and will be rolled out to users worldwide. While it may raise concerns about privacy, it remains an opt-in feature with the following safeguards:

      - It will only function if users have linked Gemini to their Search history, enabled Personalization, and activated the Web & App Activity system.

      - When Personalization is turned on in Gemini, a notification in the chat window will allow users to quickly disconnect their Search history.

      - It will transparently reveal the specifics of user data being utilized by Gemini, including saved information, previous conversations, or Search history.

      To enhance response relevance, users can also instruct Gemini to reference their prior chats. This capability has been limited to Advanced subscribers until now but will soon be available to free users globally.

      As for integrating Gemini with more applications, it has the capability to interact with various other applications—both Google's and third-party ones—through an “apps” system previously known as extensions. This offers great convenience, allowing users to accomplish tasks across different applications without having to launch each one separately.

      Google is now providing access to these apps through the Gemini 2.0 Flash Thinking Experimental model. Additionally, the selection of apps is expanding to include Google Photos and Notes, along with existing access to YouTube, Maps, Google Flights, Google Hotels, Keep, Drive, Docs, Calendar, and Gmail.

      Users can also link third-party services like WhatsApp and Spotify with their Google account to enable this apps system. Besides retrieving information and completing tasks across various applications, it also facilitates users in executing multi-step workflows.

      For instance, with a single voice command, users can direct Gemini to find a recipe on YouTube, add the ingredients to their notes, and locate a nearby grocery store. In a few weeks, Google Photos will also be included among the apps accessible to Gemini.

      “With this thinking model, Gemini can more effectively handle complex requests that involve multiple applications, as the new model can analyze the overall request, break it down into specific steps, and track its progress throughout,” explains Google.

      Furthermore, Google is increasing the context window limit to 1 million tokens for the Gemini 2.0 Flash Thinking Experimental model. AI tools like Gemini convert words into tokens, with each average English word roughly equating to 1.3 tokens.

      A larger token context window allows for a greater input size. With the expanded context window, Gemini 2.0 Flash Thinking Experimental can now process larger volumes of information and address intricate problems.