Microsoft Edge Copilot gains cross-tab AI capabilities in major update
Microsoft is retiring its standalone Copilot Mode, folding agentic features into the Browse with Copilot tool as the browser gains the ability to synthesise information from all open tabs.
Microsoft Edge is rolling out a significant update to its Copilot AI assistant, enabling the chatbot to access and synthesise information from all open browser tabs. The update introduces several new AI-driven features, including the ability to generate AI podcasts, summaries, and interactive quizzes based on browsing content. Microsoft is retiring the standalone "Copilot Mode" and integrating its agentic capabilities (such as booking reservations) into the "Browse with Copilot" tool. A redesigned new tab page is being introduced, featuring a "Journeys" function that uses AI to categorise browsing history, alongside long-term memory capabilities for Copilot.
Key new capabilities include users being able to ask Copilot to compare products across tabs, summarise open articles, and initiate conversations about the content currently displayed. A new "Study and Learn" mode allows users to convert articles into study sessions or interactive quizzes. An AI writing assistant will appear automatically when users begin typing text on a webpage. Copilot can now access browsing history to provide more relevant answers, with user permission.
The mobile app update includes a screen-sharing feature, allowing users to discuss content with Copilot while viewing their screen. Microsoft has committed to providing "clear visual cues" to indicate when Copilot is active, listening, or taking action. Copilot on desktop and mobile will feature "long-term memory" to tailor responses based on previous interactions.
The update applies to both desktop and mobile versions of the Edge browser. The new AI podcast feature is similar to functionality found in Google’s NotebookLM. The specific rollout timeline for these features (e.g., immediate availability vs. phased release) is not specified in the source material.
The exact scope of "long-term memory" and how data privacy is managed for this feature is not detailed. The precise implementation of "clear visual cues" is described but not visually verified in the text.
