“We are exploring new AI-powered capabilities that will help the cursor understand not only what it is pointing at, but also why it matters to the user,” Google DeepMind said in a blog post.

The project is designed to solve a familiar problem with today’s AI tools: most of them sit in a separate window, forcing users to copy, upload, or drag content into the assistant before getting help.

“We are aiming for the opposite: intuitive artificial intelligence that interacts with users across all the tools they already use, without interrupting their workflow,” the company said.

Four ideas behind Google’s AI cursor

Google DeepMind says the new interface is based on four principles. Their common goal is to move the burden of explaining context away from the user and toward the computer, replacing long written prompts with more natural actions such as pointing, selecting, speaking, and showing.

AI that stays inside the workflow

The first principle is continuity. AI should be available inside the apps and pages people already use, rather than forcing them to switch to a separate chat window. In Google’s prototype, the cursor can work wherever the user is active.

For example, a user could point to a PDF and ask Gemini to create a short bullet-point summary for an email. In another case, they could hover over a table of statistics and ask the system to turn the data into a pie chart.

Show instead of overexplaining

Today’s AI systems often depend on carefully written prompts. Users have to explain what they mean, where the relevant information is, and what kind of result they expect. Google DeepMind wants the cursor to reduce that friction by capturing visual and semantic context automatically.

In the experimental interface, pointing at a paragraph, an image fragment, or a piece of code may be enough for the model to understand what needs to be summarized, edited, explained, or transformed.

Making “this” and “that” useful again

Google also wants AI interaction to feel closer to everyday conversation. People naturally say things like “fix this,” “move that here,” or “what does this mean,” relying on gestures and shared context instead of long descriptions.

“An AI system that can understand this combination of context, gestures, and speech will allow users to express complex requests in a natural and concise way without relying on complicated prompts,” Google DeepMind wrote.

If the system can combine pointing, screen context, and spoken instructions, users may be able to give complex commands with short phrases instead of writing detailed prompts.

Turning pixels into actions

The final principle is about making visual content directly interactive. Rather than only tracking where the cursor is, the system should understand what is under it — a chart, a paragraph, a product image, a video frame, a form field, or a handwritten note.

“A photo of handwritten notes becomes an interactive to-do list, while a still frame from a travel video becomes a booking link for that charming restaurant,” the company said.

In practice, this would allow Gemini to treat pixels on the screen as structured objects that can be edited, summarized, moved, converted, or used as the starting point for another task.

How Google plans to use it

Google DeepMind is already bringing parts of the concept to Chrome and to a new laptop interface built for Googlebook devices.

“Starting today, instead of typing complex queries, you can use your cursor to ask Gemini in Chrome about the part of a web page you are interested in,” the company said.

In Chrome, users will be able to select several products and ask Gemini to compare them, or point to a space in a room and ask the assistant to visualize how a new sofa might look there.

Google also plans to introduce Magic Pointer on Googlebook. The feature is designed to let users call on Gemini through a simple finger movement, making AI interaction feel closer to a natural extension of the interface.

Googlebook: a new AI laptop category

Alongside the AI cursor concept, Google introduced Googlebook, a new category of laptops built around Gemini and a combined Android and ChromeOS experience.

“More than 15 years ago, we introduced Chromebook — a laptop built for a cloud-first world. Today, as we move from an operating system to an intelligent system, we see an opportunity to rethink the laptop once again,” Google said.

Googlebook devices are designed for Gemini Intelligence and the new AI cursor. They also include a “Create Widget” feature that allows users to build custom widgets with voice commands.

Gemini can search the web or connect to Google services such as Gmail and Calendar to build a personalized dashboard with relevant information, reminders, and tasks.

Because Googlebook is based on the Android technology stack, it is designed for seamless work across devices. A “Quick Access” feature will let users view, search, and insert files from their phone into a laptop workflow without manually transferring them.

The first Googlebook laptops are being developed with Acer, ASUS, Dell, HP, and Lenovo.

Gemini Intelligence expands across Android

Gemini Intelligence is Google’s broader set of AI features for Android devices. The first rollout is planned for the summer of 2026 on Samsung Galaxy and Google Pixel devices, with more device categories expected later.

“This system combines world-class hardware and innovative software to help you stay one step ahead by proactively handling tasks throughout the day, while protecting data privacy and giving users full control,” Google said in the announcement.

Google has already shown some of Gemini’s agent-style capabilities on Samsung smartphones, including ordering food and booking rides. Future versions are expected to handle more complex tasks, such as taking a photo of a travel brochure and finding similar options on Expedia.

In Chrome, Gemini will help users search, summarize, and compare information across multiple pages. In Gboard, the Rambler feature will let users dictate rough thoughts, identify the most important points, and turn them into a clean, coherent message without filler words or repetition.

Google’s AI cursor concept shows how assistants may move beyond chat windows and become part of the operating interface itself. If Gemini can reliably understand screen context, gestures, and short commands, it could make AI workflows faster and more natural across Chrome, Android, and future Googlebook devices.