Google DeepMind: AI Pointer brings Gemini-powered mouse commands to Chrome and Googlebook

Google DeepMind introduced AI Pointer on May 12, 2026 — an experimental mouse cursor powered by Gemini that allows users to point at an object on screen and speak a short command instead of typing a detailed prompt. The feature is available in Chrome immediately, while Magic Pointer for the new Googlebook laptop arrives “soon.”

How does AI Pointer change human-computer interaction?

AI Pointer reads the visual and semantic context around the cursor position and combines it with a voice command. Instead of typing “take this image, remove the background, and place it in the empty space on the left,” a user can point at the image and say “Fix this” or “Move that.” Gemini interprets the pointer position as a deictic gesture — pointing that resolves ambiguity without explanation in text.

In the Chrome version the feature covers e-commerce scenarios too: “select a few products on a page and ask to compare” or “point to where you want to visualize a new couch” directly within the web page. Everything executes without switching content to separate applications or tabs.

What is coming with the Googlebook laptop?

Magic Pointer is the name of the deeply integrated Googlebook implementation that Google plans to launch soon. DeepMind is targeting seamless AI assistance across multiple applications “without workflow interruption” — suggesting system-level access, not merely a browser plugin.

Two experimental demo applications are already available in Google AI Studio: an image editing tool and a map-based location finder. Both share the “point and speak” pattern and serve as a playground for developers who want to explore how the Gemini API can connect visual context and voice commands in their own applications.

Position in the broader AI ecosystem

AI Pointer follows a wave of 2026 products that are reorganizing classic UI elements around large language models — Google AI Mode in Search, Microsoft Copilot in Windows, Anthropic Computer Use in the Claude API. The mouse cursor was previously dumb and behaved identically in every application; DeepMind is turning it into a context-aware input device that understands the content beneath the cursor.