At AI Chatbot Wars, Google believes that the key to keeping users is providing content that you don't get anywhere else, like the answers shaped by internet habits.
On Thursday, the company announced Gemini. Personalization is a new “experimental feature” in the Gemini Chatbot app, allowing Gemini to use other Google apps and services to provide customized responses. According to Gemini Product Director Dave Citron, Gemini with personalization allows users to tap on user activity and preferences across Google's product ecosystem to provide answers tailored to their queries.
“All of these updates are designed to make Gemini feel like a tool, not a natural extension for you, and we look forward to your needs with truly personalized support,” Citron wrote in a blog post provided to TechCrunch. “Early testers discovered Gemini with personalizations that help them brainstorm and get personalized recommendations.”
Image credit: Google
Gemini will integrate with Google Search before expanding to additional Google services such as Google Photos and YouTube, and will integrate with Google Search within the next few months. Openai recently deployed the ability to edit code directly in apps that support MacOS ChatGpt, while Amazon prepares to begin rethinking Alexa's “agent” versions.
Citron said Gemini with personalization is powered by Google's experimental Gemini 2.0 Flash Thinking Experimental AI model. There are narrow questions that are told by likes and dislikes, such as, “Where should I take a holiday this summer?” And then Citron continued, “Can you suggest I learn as a new hobby?”
“For example, you can ask Gemini for restaurant recommendations and refer to recent food-related searches,” he said.
Image credit: Google
If this all sounds like a privacy nightmare, it might be. It's not difficult to imagine a scenario in which Gemini accidentally airs someone's sensitive information.
That's probably why Google is creating Gemini with personalized opt-in, and it excludes users under the age of 18. Gemini asks for permission before connecting to Google Search History or other apps, indicating which data sources were used to customize the bot's response.
“When using personalization experiments, Gemini displays a clear banner with a link to easily disconnect your search history,” Citron said. “Gemini only accesses your search history when you select Gemini for personalization, when Gemini gives permission to connect to your search history, and if you are using Web & App activity.”
Image credit: Google
With personalization, Gemini will be deployed to Gemini users on the web (excluding Google Workspace for Education for Education customers) starting Thursday in the APP model drop-down menu, and then “gradually” become mobile. Except for the European Economic Area, Switzerland and the UK, Citron said it will be available in more than 40 languages, the “majority” of the country.
Citron has shown that this feature may not be free forever.
“Future use restrictions may apply,” he wrote in a blog post. “We will continue to collect user feedback on the most useful applications of this feature.”
New models, connectors, etc.
With the addition of incentives to stick to Gemini, Google has announced an updated model, research features, and app connector for the platform.
Subscribers of Gemini Advanced, a Google $20/month premium subscription, can now use the standalone version of 2.0 Flash Thinking Experimental, which supports file attachments. Integrate with apps such as Google Calendar, Notes, Tasks, and more. And a context window for 1 million tokens. A “context window” refers to text that the model can always consider. One million tokens amount to about 750,000 words.
Google said this latest version of 2.0 Flash Thinking Experimental is faster and more efficient than the model it is replacing, and can better handle prompts that include multiple apps, such as “find simple cookie recipes on YouTube, add ingredients to your shopping list, and find a grocery store that's operating nearby.”
Perhaps in response to the pressure from Openai and the newly launched tools for in-depth research, Google is bolstering deep research. This is a Gemini feature that searches the entire web and compiles reports on themes. Deep Research now publishes its “thinking” step, and with 2.0 flash thinking being used experimentally as the default model, it should generate more “detailed” and “insightful” and “high quality” reports.
Deep Research is free to try it out on all Gemini users, and Google is increasing the use limits for advanced Gemini customers.
Free Gemini users also get Gems, a Google topic-centric, customizable chatbot within Gemini. And in the coming weeks, all Gemini users can interact with Google Photos to look up photos from recent trips, for example. Google said.