Google LLC today debuted a tool called Personal Intelligence that will enable consumers to tailor its Gemini chatbot to their preferences.

The tool will initially be available to a limited number of paying users in the U.S. Google plans to expand access over time.

Consumers can enable Personal Intelligence through a new option in the Gemini settings menu. The tool gives the chatbot access to information that the user keeps in other Google services such as Gmail, Google Photos and YouTube. Gemini draws on that data to refine its prompt responses.

If the chatbot receives a request for recipe suggestions, it might analyze restaurant reservations in the user’s inbox and customize its output accordingly. Gemini can also make use of multimodal data. For example, it might pull photos of the user’s car from Google Photos to answer maintenance-related questions. 

“We needed new tires for our 2019 Honda minivan two weeks ago,” Google executive Josh Woodward wrote in a blog post. “I asked Gemini. These days any chatbot can find these tire specs, but Gemini went further. It suggested different options: one for daily driving and another for all-weather conditions, referencing our family road trips to Oklahoma found in Google Photos.”

Gemini lists the personal data sources that it uses in prompt responses to help consumers verify the information. According to the company, the chatbot limits its use of personal data that pertains to sensitive topics such as the user’s health. However, Gemini can incorporate that information into answers if it’s specifically asked to do so.

Google equipped Personal Intelligence with several privacy controls. The feature is disabled by default and users who enable it can customize which Google services it accesses. Additionally, a feature called temporary chats makes it possible to turn off personalization for specific chatbot sessions.

Google will use Gemini’s prompt responses to train its artificial intelligence models, but only after filtering any sensitive data they contain. For example, Personal Intelligence can remove the user’s license plate number from an answer to a car-related question. Users can also go a step further by deleting Gemini’s chat history.

Gemini 3, the large language model series that powers the chatbot, can process prompts with up to 1 million tokens. The challenge Google encountered while developing Personal Intelligence is that the user file repositories the tool is asked to analyze often contain more than 1 million tokens. According to the company, its engineers overcame that obstacle using a method called context packing. The technique enables Gemini to find and use only the specific subset of a user dataset that is relevant to a given prompt.

Personal Intelligence will roll out to early adopters in the U.S over the next week. Initially, accessing it will require a paid Google AI Pro or Google AI Ultra subscription. Over time, the company plans to expand the availability of the tool to free accounts and international markets. It will also embed Personal Intelligence in Google Search’s AI Mode. 

Image: Google