Microsoft launches Copilot Vision preview in Edge to browse the web with you

Staring at dozens of open browser tabs while comparing flights, holiday gifts, or dense research papers can feel exhausting. Microsoft’s latest bet is that browsing shouldn’t be a solo chore, so it gave its Edge browser a pair of eyes and a voice to tag along for the ride.

Microsoft has officially rolled out a preview of Copilot Vision inside its Edge browser. First revealed in October, the feature is making its debut through Copilot Labs, Microsoft’s new testing ground for experimental AI tools. It is currently rolling out to a limited group of Copilot Pro subscribers in the United States.

An AI friend sitting at the bottom of Edge

Unlike traditional chatbots that require you to copy-paste text or manually upload screenshots, Copilot Vision sits at the bottom of Edge and sees the webpage alongside you. Once enabled, you can talk to it in real time about whatever is on your screen.

Demos show the assistant helping users find the right vacation rental, deciphering handwritten recipes from old photos, and pointing out product details while shopping online. Microsoft AI chief Mustafa Suleyman emphasized that the goal is to build a genuine companion that can see what you see and hear what you hear.

Interestingly, Microsoft is leaning heavily into personality. Suleyman noted that the voice is smooth, highly interruptible, and dynamic enough to mirror your emotional state. If you are excited, it picks up the pace; if you speak slowly, it matches your vibe. It can even push back and deliver a bit of sass occasionally, because, as Suleyman put it, nobody wants a sycophantic AI that blindly agrees with everything.

Walking the privacy tightrope after Recall

Allowing an AI model to watch your screen naturally rings alarm bells, especially after the backlash Microsoft faced over its Windows Recall feature. This time around, the company is treading far more carefully.

Copilot Vision is strictly opt-in, requiring explicit permission before it can view any page. Sessions are ephemeral: the moment you end the chat, the webpage content and conversation data are wiped clean, with only Copilot’s own responses logged for safety monitoring.

Microsoft is also playing it safe with publishers. For now, Vision only works on a select whitelist of approved websites. It is explicitly programmed to ignore paywalled content, respects crawler block tags, and does not store or use publisher content to train its models. Microsoft plans to gradually expand website support before pushing the feature to general availability sometime early next year.

Why it matters

We think Copilot Vision is a clever pivot in Microsoft’s consumer AI strategy. By integrating screen awareness directly into the browser, Microsoft is trying to turn AI from a destination you visit into ambient scaffolding that sits inside your daily workflow.

For users, having an assistant that understands page context without tedious prompt-engineering is genuinely helpful, particularly for research and shopping. However, at $20 a month for Copilot Pro and with access restricted to a handful of whitelisted sites, it is not something ordinary users need to rush out and buy today. It makes far more sense to wait for broader site support.

For the broader industry, this move puts pressure on competitors like Google. While OpenAI focuses on standalone apps and voice modes, Microsoft is weaponizing its Edge browser distribution to make real-time vision native. If Microsoft can prove that people actually want a vocal co-pilot watching their web traffic without triggering privacy revolts, expect every major browser vendor to copy the playbook fast.