Copilot
Extract structured data from the Microsoft Copilot web interface - citations and links across five distinct response modes - as clean JSON.
search for cited web answers, reasoning for deep analysis, and study for structured educational content.Overview
Use this endpoint to track brand mentions in Microsoft’s AI ecosystem, run structured research across different answer styles, and monitor how Copilot responds to industry queries. Thestudy mode is particularly useful for generating structured content briefs.
Unique features
- Five modes: Switch between web search, smart reasoning, chat, deep reasoning, and study-guide formats
- Named citations: Cited sources with title and URL via
citations[] - Raw links: All outbound links surfaced in the response via
links[]
Modes
Example request
Response schema
Includes common response fields plus:Common questions
What is the difference between citations and links?
citations are the named sources Copilot explicitly attributes in its answer (with titles). links are all outbound URLs surfaced in the response - including links within the answer body that may not be formal citations.
Which mode is best for brand monitoring?
search mode - it’s web-grounded and always includes citations, making it the most reliable for tracking which sources Copilot recommends for any industry query.
When should I use reasoning mode?
For complex, multi-part questions where you need analytical depth - competitive analysis, strategy questions, or topics requiring Copilot to weigh multiple factors. Responses take longer but are more thorough.
Are JP and TW supported?
No. Japan and Taiwan are not currently supported. Use any other ISO 3166-1 alpha-2 country code.Authorizations
Your ScrapeLLM API key passed in the request header.
Query Parameters
The prompt to send to Copilot. Maximum 4,000 characters.
1 - 4000ISO 3166-1 alpha-2 country code. Routes the request through infrastructure in
that region for localised responses. JP and TW are not supported. Defaults to US.
2Copilot response mode:
search— web-grounded answer with cited sources (default)smart— balanced blend of web search and reasoningchat— quick direct reply, no web searchreasoning— slower, deeper analysis for complex questionsstudy— structured study guide format
search, smart, chat, reasoning, study Skip the response cache and always fetch a fresh result.
Maximum seconds to wait for Copilot to respond. Must be between 10 and 600.
10 <= x <= 600Response
Scrape completed successfully.
Always "copilot".
"copilot"
"done" on success.
"done"
Unique identifier for this scrape job.
"job_abc123"
The prompt that was submitted.
"What brands do marketers recommend for email automation?"
The country the request was routed through.
"US"
The Copilot mode used for this request.
search, smart, chat, reasoning, study "search"
Plain-text response from Copilot.
"Marketers commonly recommend Mailchimp, HubSpot, and Klaviyo..."
Credits consumed by this request. Always 3 for Copilot.
3
Total end-to-end request duration in milliseconds.
9102.4
true if this response was served from cache.
false
Markdown-formatted response from Copilot.
"Marketers commonly recommend **Mailchimp**, **HubSpot**, and **Klaviyo**..."
Sources explicitly cited by Copilot in its answer, each with a title and URL.
All outbound URLs returned by Copilot (superset of citations).
Hosted .md URL for this response. Expires after 24 hours.
"https://scrapellm.com/md/job_abc123.md"