Google's new feature adds an AI chatbot to Search
What's the story
Google is about to transform its search experience with AI Mode. The feature, powered by Gemini 2.0 model, has been trained by Google to browse and organize information on the internet.
The results will appear in a conversational interface that allows follow-up questions.
Robby Stein, Google's Vice President of Product for Search, said "the goal and vision for [AI Mode] is to empower people to search for anything effortlessly."
User feedback
AI Mode: A response to user demand
The creation of AI Mode was driven by user feedback on Google's earlier feature, AI Overviews.
Stein disclosed that after rolling out this feature to over a billion users, many wanted more AI-generated summaries in their search results.
Some even started putting "AI" at the end of their searches as a prompt for Google to respond with an AI Overview.
Advanced technology
Gemini 2.0 model: The power behind AI Mode
The Gemini 2.0 model powers AI Mode. It employs a "query fan-out" technique to perform multiple searches at the same time on different topics and data sources.
Plus, it also utilizes Google's proprietary information systems, such as the Knowledge Graph, to enrich its online results with additional information from internal sources.
User interface
Aim to deliver a rich visual experience
Google is currently rolling out a version of AI Mode for its Google One AI Premium subscribers.
Though not all planned interface elements are implemented yet, the idea is to provide a rich visual experience.
For example, if you ask AI Mode to compare different sleep trackers, you might soon see Gemini generate a comparison chart.
The tool will also display sources more prominently than before and default to showing regular search results when it can't produce an authoritative answer.
Usage patterns
User interaction with AI Mode reveals new insights
Testing of AI Mode has also indicated that users are interacting with it in a different way than a traditional search engine.
Stein noted that testers' queries were about twice as long and they asked follow-up questions about a quarter of the time.
This indicates users are bringing more complex questions to this platform and using it for clarification, learning, and deeper exploration.