Today, Google has revealed and begun testing a new “AI Mode” for its Search feature. This “early experiment” allows users to pose more intricate inquiries that will be addressed by AI-generated answers occupying the full screen.
Based on user testing, Google discovered that individuals appreciate AI Overviews. This leads to more follow-up searches and prolonged, distinct queries, as users realize that the Search can tackle a wider range of questions with the aid of AI.
The company highlights that “power users” are eager for AI-driven responses for diverse search types. In fact, Google has observed users adding “AI” to their queries, which prompted the introduction of AI Mode. Unlike AI Overviews, this feature offers “advanced reasoning, critical thinking, and multimodal capabilities.”
AI Mode ensures users receive direct answers while also permitting them to pose more “nuanced,” multi-part questions that might have required several queries in the past.
On both desktop and mobile devices, users can enter a query into Google as usual and then click the new “AI Mode” button next to the filters for All, Images, Videos, etc. A “Go deeper” shortcut is also available at the bottom of AI Overviews, and dedicated entry points to AI Mode enable users to bypass typical results and engage directly with the new feature. The query appears at the top of the page, with the generated response displayed below it.
Instead of a conventional Search bar, a chat-style “Ask a follow up…” field is located at the bottom. On mobile, you have the option to upload images and voice your query to AI Mode, though text output is currently the only response format. A thread history allows you to view previous searches.
For instance, if the query is: “how do migrating birds know where to go,” AI Mode will conceptualize (or devise a plan), conduct multiple searches simultaneously, and organize the findings. On mobile, a carousel showcases the sites used as sources before providing a succinct answer. Subsequently, another carousel displays articles, with specific details (in this example) presented afterwards.
AI Mode is fueled by a tailored version of Gemini 2.0, which has access to real-time data and information from the real world, including the Knowledge Graph and Shopping Graph covering billions of products. Google conducts multiple related searches concurrently and consolidates those results. This technique, termed “query fan-out,” allows for the discovery of information across subtopics and numerous data sources before synthesizing it.
The model is designed to identify and substantiate its claims. If it lacks sufficient information, users will be directed to web results. Responses are intended to be factual and objective, devoid of personal opinions or bias.
In another example, when someone asks: “when is the best time this week to schedule an outdoor engagement photoshoot in the Boston public garden,” Google offers a specific date based on optimal weather conditions (using real-time data) and suggests the golden hour, specifying the sunset time.
In the final example, the differences between AI Mode on mobile and desktop are clear when asking: “what’s the difference in sleep tracking features between a smart ring, smartwatch and tracking mat.”
The response comes in the form of a comparison table (which is not yet active but will be available soon), and users can ask follow-up questions at any time, such as “what happens to your heart rate during deep sleep.”
Although the traditional display of 10 blue links is absent, users will find inline sources as well as a side card feature on desktops. Looking ahead, Google plans to introduce a rich carousel of links. The aim of AI Mode is to connect users with the web, as research indicates they fundamentally wish to verify information and view various sources alongside AI-generated responses. The underlying model is configured to understand how to reference sources.
From initial testing, AI Mode queries tend to be twice the length of standard queries, with users following up approximately one-quarter of the time.
On a broader scale, Google observes that users recognize when to use standard Search versus AI Mode. These two experiences are designed to coexist, with Google currently enhancing AI Overviews with Gemini 2.0 in the US for complex coding, advanced mathematics, and multimodal (like Lens) queries. This list will grow over time, while other queries will continue using the existing model. Google asserts that Gemini 2.0 enables “faster and higher quality” responses, with AI Overviews appearing “more frequently for these types of searches.” Concurrently, Google is expanding AI Overviews globally, and users need not be signed in.
Beginning today, Google One AI Premium ($19.99 per month) subscribers in the US will receive invitations (via email) to opt into AI Mode within Search Labs. You can also join the waitlist here. Over time, the objective is to allow anyone to access this experience, but current testing is focused on power users.
