To commemorate its 50th anniversary, Microsoft is enhancing its AI-powered Copilot chatbot with several new features.

Copilot now has the capability to interact with “most websites,” enabling it to perform tasks such as booking tickets, making restaurant reservations, and more, according to Microsoft. The bot has also gained the ability to recall specific details about users, similar to OpenAI’s ChatGPT, including their favorite foods and movies. Additionally, it can analyze real-time video from a user’s phone, providing answers within the context of what it “sees.”

These updates coincide with Microsoft’s reported plans to revamp Copilot, which has traditionally relied on AI models from OpenAI, by incorporating more of its own in-house technology. Copilot has often lagged behind its competitors, ChatGPT and Google’s Gemini, which have recently accelerated their feature rollout pace.

As of the latest update, Copilot can complete tasks on the web in a manner similar to “agentic” tools like OpenAI’s Operator. Microsoft has partnered with several companies, including 1-800-Flowers.com, Booking.com, and Expedia, to enable day-one compatibility. By entering a prompt, such as “send a bouquet to my partner,” Copilot will attempt to fulfill the request.

Inspired by search engine Perplexity’s shopping feature, Copilot can now track online deals and notify users when prices drop or sales occur, providing a link to make a purchase.

However, the effectiveness of Copilot in performing various tasks is unclear, as Microsoft has provided limited details on its functionality and has not published data on potential areas of struggle or the need for human intervention.

It is possible for websites to block Copilot, similar to OpenAI’s Operator, if they are concerned about the potential impact on their ad revenue, such as fewer direct app visits.

Fortunately, Copilot’s other new features are more straightforward and less controversial.

The updated Copilot can generate podcasts similar to Google’s NotebookLM Audio Overviews. Given a website, study, or other source, Copilot will create a dialogue between two synthetic hosts. Users can interrupt the hosts at any point to ask a question, and they will respond accordingly.

On Android and iOS devices, Copilot can now analyze the view from the phone’s camera or photo gallery, answering questions about the content (e.g., “What’s this flower?”). On Windows, the revamped Copilot app can view the desktop screen to search, change settings, organize files, and more, with the update rolling out first to Windows Insider program members next week.

It is hoped that reasonable safeguards are in place to prevent Copilot from accessing private files or causing desktop errors. However, information on this was limited prior to press time.

Additionally, Copilot has introduced a new project-consolidating Pages function, drawing inspiration from ChatGPT Canvas and Anthropic’s Claude Artifacts tool. Pages organizes notes and research into a canvas that Copilot can help structure and turn into a document.

Complementing Pages, Copilot’s new Deep Research feature finds, analyzes, and combines information from online sources, documents, and images to answer complex queries, similar to ChatGPT deep research and Gemini’s Deep Research.

Lastly, as mentioned earlier, Copilot can now remember more about users. Microsoft states that the bot will note preferences as users interact with it, offering tailored solutions, proactive suggestions, and reminders.

For users concerned about the prospect of a chatbot remembering intimate details from past conversations, there is an option to delete individual “memories” or opt out entirely, according to Microsoft.

“Copilot provides users with control through the user dashboard and the option to choose which types of information it remembers about them or to opt out entirely,” Microsoft wrote in a blog post provided to TechCrunch. “Users remain in control.”