Google adds limited chat customization to Gemini and trails Anthropic in memory features

Google catches up with the competition Anthropic Openai is slowly adding customization, personalization, and giving users more control over which data to reference in its Gemini app.

Personalization in chat platforms and data control make it easier for individual and enterprise users alike to converse with chatbots and retain preferences. This is especially important for projects that are ongoing in the enterprise space as chatbots must remember details like company branding or voice. Google, unlike its competitors, has chosen to roll out these features slowly and won’t allow users the option to edit or delete their preferences.

Google will first make “Personal Context”a default setting for Gemini 2.5 Pro, in select countries. This feature allows it to “learn” from past conversations, and “provide relevant and tailored answers.” The company plans on expanding the feature to Flash 2.5 in the next few week.

Previous versions of the app put the burden on customers to point the model to a specific chat to source preferences, for example, by mentioning an earlier conversation. Users can disable Personal Context anytime. Michael Siliski is the senior director of Product Management at Gemini, and he said that the rollout was part of plans to personalize the app. Siliski wrote in a blog that “at I/O we presented our vision for the Gemini App: to create an AI Assistant that learns from you and understands your needs, not one that just responds the same as it would to anyone else’s request.”

Gemini apps currently save chats up to 72 hours when the save activity option has been toggled off. Other activity can be automatically deleted in intervals of 3, 18 or 36 months.

Temporary chats and data control

Temporary Chats and additional customer data controls are also new features that will be coming to the Gemini App.

Temporary Chat is a feature that was also launched on ChatGPT last April. It allows for users to have a one-off conversation. These chats won’t affect future ones, and they won’t be used to personalize or train AI models. Google announced that it would be introducing additional data controls. The feature is disabled by default and allows users to prevent data from being used for future Google model training.

When this setting is enabled, a sample from your future uploads can be used to improve Google services. If you don’t want your data to be used in this way, turn off this setting or use Temporary chats. Silisky explained that if your Gemini Apps Activity Setting is currently off, the Keep Activity setting will also remain off. You can turn it back on at any time.

Google says this is an extension of an earlier update which allowed users to select which audio, videos and screens they could share with Gemini.

Memory and chatbots.

Google Gemini updates arrive a year after its largest competitors introduced similar features.

ChatGPT introduced temporary chat, chat histories and memory in 2024. OpenAI updated the capabilities in April this year. Now ChatGPT is able to reference all previous conversations.

Anthropic launched Styles in November of 2024. This allows Claude users the ability to customize how their model interacts with them. Anthropic released an update for Claude earlier this week. Reference all conversationsand not just those specified by users.

Although Google introduced personalization in Gemini 2.0, this model could only refer to previous conversations if the user prompted it.

Memory and personalization continue to be battlegrounds in the AI arms races as users expect chat platforms to “just” know them or their brands. It provides context for ongoing projects and eliminates the need for repeating instructions.