The Researcher agent in Microsoft 365 Copilot helps you gather, analyze, and summarize information from the web and your work content. Researcher supports multiple AI models, including GPT models from OpenAI and Claude models from Anthropic, giving you flexibility in how research reports are generated and refined. Depending on the option you select, Researcher will perform deep reasoning and generate multiple research reports for the same query or let you work with a specific model directly. 

Your admin must first allow access to Anthropic AI models in the Microsoft 365 admin center before you can use Claude in Researcher. For more information, see Anthropic as a subprocessor for Microsoft Online Services. 

For more tips on using the Researcher agent, see Get started with Researcher.  

You must have a Microsoft 365 Copilot license to use model choice in the Microsoft 365 Copilot app (desktop and web app only).

Note: You need to be part of the Frontier program to use the new features in Auto and Model Council. Frontier lets you try the latest model innovation and give feedback before these experiences are generally available. Frontier includes early access to experimental features, which means features may change as Microsoft improves them. 

Critique model

The default model when Anthropic is enabled uses Critique automatically. Critique is a built-in review step in Researcher that checks and improves a draft report before it’s delivered. With the Critique model, a report is generated by GPT and is given a second reasoning pass with Claude to strengthen structure and completeness. Critique model prioritizes reputable sources, and ensure key statements are grounded in clear citation to make the final report more reliable and easier to trust. 

Use Critique model 

  1. Sign in to the Microsoft 365 Copilot app and select Researcher under Agents in Chat.  Auto should already be selected for you.

    Screenshot: Image showing select Auto on Researcher agent page

Model Council

Model Council is Microsoft 365 Researcher’s multi-model comparison mode. With Model Council, Researcher runs the same question through multiple deep-reasoning research agents (GPT and Claude) at the same time. Model Council mode preserves each model’s full standalone report and then adds a summary that highlights where the models agree, where they diverge, and what each uniquely contributes. You can then quickly compare perspectives and choose the output (or combine insights) with higher confidence. 

Use Model Council

  1. Sign in to the Microsoft 365 Copilot app and select Researcher under Agents in Chat.

  2. Select Model Council from the model dropdown and enter your prompt.

    Screenshot: Image showing selection of Model Council model

GPT model

Use this option if you are only interested in the report from a GPT model.  

Use GPT model

  1. Sign in to the Microsoft 365 Copilot app and select Researcher under Agents in Chat.

  2. Select GPT from the model dropdown and enter your prompt.

    Screenshot: Image showing selection of GPT model

Claude model

Use this option if you are only interested in a report from a Claude model. 

Use Claude model 

  1. Sign in to the Microsoft 365 Copilot app and select Researcher under Agents in Chat.

  2. Select Claude from the model dropdown and enter your prompt.

    Screenshot: Image showing selection of Claude model

Need more help?

Want more options?

Explore subscription benefits, browse training courses, learn how to secure your device, and more.