Decision briefs
Research a market, vendor or strategic question and use the second model pass to challenge incomplete coverage or weakly grounded claims.
Independent tool overview
Critique at a glance
Critique is a built-in multi-model review mode for Microsoft's Researcher agent, not a standalone app. GPT generates a research report and Claude performs a second reasoning pass intended to improve its sourcing, completeness, grounding and presentation before delivery.
Visit the official Critique site ↗
- Product type
- Multi-model review mode inside Microsoft Researcher
- Model flow
- GPT drafts; Claude reviews and refines
- Availability
- Frontier early access as of August 31, 2026
- Admin requirement
- Anthropic models must be allowed for organizational accounts
- Best for
- Source-cited, multi-step research using web and Microsoft 365 work context
Overview
What Critique is
Microsoft Critique separates report generation from evaluation. Researcher first uses GPT to plan, retrieve sources and draft the answer; Claude then reviews that draft against a rubric and produces an improved final report. This is a sequential reviewer workflow, unlike Model Council, which preserves multiple parallel reports for side-by-side comparison.
Critique operates inside the Microsoft Copilot app's Researcher agent. Researcher can combine public web sources with work content the user is already permitted to access, including files, email, meetings and chats, and returns a structured report with citations.
As of August 31, 2026, Microsoft's support documentation classifies Critique as a Frontier early-access feature. It requires a Microsoft Copilot license for model choice, and an organization administrator must allow Anthropic models before Critique can use Claude.
Use cases
Who Critique is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Work-grounded research
Combine public evidence with files, emails, meetings and chats already accessible through the user's Microsoft 365 permissions.
High-visibility reports
Create a structured, cited first draft for a document or presentation where source quality and clear organization matter.
Capabilities
Core Critique features
Separate generator and reviewer
GPT produces the initial research report, then Claude reviews and refines it instead of asking one model to generate and grade its own work.
Source-reliability review
The review rubric prioritizes reputable, authoritative and context-appropriate evidence.
Completeness review
Critique looks for missing angles and whether the report fully addresses the user's actual request.
Citation grounding
The reviewer is instructed to anchor important claims to clear citations and improve the use of existing evidence.
Web and work context
Researcher can use web results and, for organizational users, Microsoft 365 content the requesting user is authorized to access.
Related model choices
Eligible users can also select GPT-only, Claude-only or Model Council modes when they want a single-model answer or parallel comparison.
Process
How the Critique workflow works
Step 1
Confirm access
Use an eligible Microsoft Copilot license, join the Frontier program and, for an organization, have an administrator enable Anthropic models.
Step 2
Open Researcher
Sign in to the Microsoft Copilot app on desktop or the web and select Researcher under Agents.
Step 3
Select Critique
Choose Critique in the model dropdown. Microsoft also documents Critique as the automatic default in Auto when Anthropic is enabled.
Step 4
Define the research task
State the decision, required scope, date range, preferred source types and desired output. Answer any clarifying questions Researcher asks.
Step 5
Audit the final report
Open material citations, check sensitive conclusions against primary sources and edit or export the report before sharing it.
Cost
Critique pricing and free plan
Microsoft does not sell Critique separately. Access is part of eligible Microsoft 365 Premium or Microsoft 365 Copilot licensing, subject to Frontier enrollment, model availability and administrator policy. U.S. list prices reviewed on August 31, 2026 were $19.99 per month or $199.99 per year for Microsoft 365 Premium and $30 per user per month paid annually for the commercial Microsoft 365 Copilot add-on; the commercial add-on also requires a qualifying Microsoft 365 plan.
Microsoft 365 Premium
$19.99/month or $199.99/year
Consumer subscription with access to advanced AI agents including Researcher; AI benefits apply to the subscription owner.
- Includes the Microsoft 365 apps and up to 6 TB total family storage
- Critique remains subject to Frontier and feature availability
- Promotional prices may differ from the regular U.S. list price
Microsoft 365 Copilot
$30/user/month, paid yearly
Commercial Copilot add-on with work-grounded Researcher and model choice.
- Requires a separate qualifying Microsoft 365 license
- Administrator must allow Anthropic as a subprocessor for Critique
- Desktop and web Microsoft Copilot apps support model choice
Business Premium with Copilot
$32/user/month, paid yearly
Bundled U.S. business plan combining Microsoft 365 Business Premium and Copilot.
- Annual auto-renewing subscription
- Includes productivity, security and work-grounded AI capabilities
- Availability and taxes vary by market
Pricing checked . Check current pricing at the source ↗
Assessment
Critique strengths and limitations
Where it stands out
- A second model reviews the draft, creating a more meaningful quality-control step than a single model simply checking itself.
- The rubric explicitly targets source reliability, coverage gaps, evidence grounding and report presentation.
- Researcher can work with both the public web and permission-respecting Microsoft 365 context.
- The result stays inside the Microsoft Copilot workflow and can be edited, shared or used to start a document or presentation.
What to consider
- Critique is not a standalone service and is still a Frontier early-access feature, so access and behavior can change.
- Organizations must approve Anthropic as a subprocessor; without that setting, the Claude review pass is unavailable.
- Two-model processing takes longer than normal Copilot Chat and should be reserved for work that benefits from deeper research.
- A reviewer model can still miss false claims, misread a source or preserve a bad citation; the workflow does not replace human verification.
- Researcher can surface information from content a user is permitted to access, so organizations should audit oversharing and existing Microsoft 365 permissions before broad rollout.
- Microsoft's published DRACO improvements are Microsoft-run evaluations of its own system, not a guarantee for every research topic.
Compare
Critique alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Data Analysis
Perplexity
Perplexity is a simpler web-first research option when Microsoft 365 work grounding and administrative controls are not required.
Explore Perplexity →Business Operations
ChatGPT
ChatGPT offers deep-research capabilities and broader general-purpose tools outside the Microsoft 365 environment.
Explore ChatGPT →Consumer
Deep Research Max
Google Deep Research Max is an alternative for long-form, source-backed research in the Gemini ecosystem.
Explore Deep Research Max →Questions
Critique FAQs
Is Microsoft Critique a standalone app?
No. Critique is a model mode inside the Researcher agent in the Microsoft Copilot app.
Which models does Critique use?
Microsoft documents a GPT model generating the initial report and a Claude model from Anthropic performing the second review and refinement pass.
How is Critique different from Model Council?
Critique uses one report pipeline: GPT drafts and Claude reviews. Model Council runs multiple research agents in parallel, preserves their separate reports and adds a comparison of agreements and disagreements.
Who can access Critique?
As of August 31, 2026, Microsoft says users need an eligible Microsoft Copilot license and Frontier access. Organizational administrators must also enable Anthropic models.
Does Critique guarantee accurate research?
No. It adds a useful independent review pass, but both models can still make errors. Verify important claims and citations directly, especially for legal, medical, financial or strategic decisions.
Does Critique cost extra?
There is no separate Critique fee. It is included with eligible Copilot licensing and feature access, but commercial users also need a qualifying Microsoft 365 plan.
Bottom line
Our Critique verdict
Critique is a thoughtful upgrade for teams already using Microsoft 365 Copilot: assigning the quality check to a different model can improve report discipline, and work-data grounding makes it unusually useful for internal research. Its Frontier status, Anthropic approval requirement and continued need for human source verification make it a controlled pilot rather than an automatic replacement for expert review.
Visit Critique website ↗Similar Miscellaneous Tools
View all Miscellaneous AI tools →
TADA by Hume AI
TADA - Hume AI's open-source TTS model that syncs text and audio one-to-one for zero-hallucination speech
MAI-Transcribe-1 (Deprecated)
MAI-Transcribe-1 - Microsoft's speech-to-text model with best-in-class accuracy across 25 languages
Gemini Embedding 2
Gemini Embedding 2 - Google's multimodal model capable of searching across text, images, video, and audio at once
Google Edge Eloquent
Google Edge Eloquent - Google AI Edge Eloquent — Free voice dictation app that turns messy speech into polished text, runs fully offline, no subscription, no usage caps
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.