Google Gemini Adds Personalization, Trails Rivals in Memory Features
Google is incrementally enhancing its Gemini application with new personalization and data control features, aiming to provide users with more tailored interactions. This rollout positions Google in a race to catch up with rivals like Anthropic and OpenAI, who have offered similar capabilities for some time.
One of the key additions is “Personal Context,” a setting designed to allow Gemini to learn from past conversations and deliver more relevant and customized responses. This feature will initially be a default setting for Gemini 2.5 Pro users in select countries, with plans to extend it to Gemini 2.5 Flash in the coming weeks. Unlike some competitors, Google will not immediately allow users to edit or delete these learned preferences. Previously, users had to explicitly direct the model to specific chats to reference earlier preferences, for example, by mentioning a prior conversation. While “Personal Context” is enabled by default, users retain the option to disable it at any time.
Michael Siliski, senior director of Product Management for the Gemini app, articulated the broader vision behind these updates. He stated that the goal for the Gemini app is to evolve into an AI assistant that genuinely learns and understands its users, moving beyond a model that delivers identical responses to everyone. This deeper understanding is particularly critical for enterprise users, where chatbots need to retain specific details such as company branding or communication style for ongoing projects.
Beyond personalization, Google is also introducing “Temporary Chat,” a feature enabling one-off conversations that will not influence future interactions, nor will they be used for personalization or AI model training. This functionality mirrors a similar feature introduced by ChatGPT in April of the previous year. Additionally, Google is rolling out enhanced customer data controls. This setting, which is off by default, allows users to prevent their data from being used in future Google model training. If enabled, a sample of user uploads will contribute to improving Google services broadly. This builds on an earlier update that gave users more control over sharing audio, video, and screen content with Gemini. Currently, Gemini apps save chat activity for up to 72 hours if the save activity option is disabled, and can automatically delete other activity at intervals of three, 18, or 36 months.
These updates from Google arrive approximately a year after its primary competitors launched comparable features. OpenAI’s ChatGPT, for instance, introduced temporary chat, chat history, and memory capabilities in 2024, further enhancing these in April of this year to allow the model to reference all past conversations. Similarly, Anthropic launched “Styles” for its Claude model in November 2024, enabling users to customize the model’s interaction style, and recently updated Claude to reference all previous conversations without requiring explicit user prompts. While Google did introduce a form of personalization to Gemini 2.0, that model still required users to prompt it to recall earlier conversations.
The ability for AI chat platforms to “remember” and understand user preferences without constant prompting has become a crucial battleground in the rapidly evolving AI landscape. This “memory” provides essential context, eliminating the need for users to repeatedly provide instructions or details for ongoing projects, and is a key differentiator in the race to deliver truly intelligent and intuitive AI assistants.