Microsoft Copilot Scraper - Answers, Citations & Brand Tracking

Microsoft Copilot Scraper sends your questions to Microsoft Copilot and returns every answer as clean, structured data: the full answer text, every cited source, inline citations, tables, product and place cards, research sources, and brand mentions. Pick one of five Copilot modes and a country. No Microsoft login, no account, no API key.

New here? Click Try for free, keep the two prefilled questions, and press Start. In under a minute you will see real Copilot answers in the dataset, with their sources and cards.

What you get

• The full Copilot answer as Markdown, including its lists and tables, exactly as Copilot wrote it.
• Every cited source with title, URL and domain, numbered in the order Copilot cites them, plus the inline citations that tie spans of the answer to their source.
• Research sources Copilot read but did not cite inline, kept separate so you never mistake them for citation wins.
• Product cards when Copilot answers a shopping question: name, price, rating, seller, brand, pros and cons, specifications and the offers behind each one.
• Place cards for local questions: name, address, coordinates, phone, category, rating and review count.
• Brand tracking: tell it your brand and competitors, and every answer reports whether each is mentioned, how often, its rank, and whether its site is cited. A run-level summary rolls this up into mention rate, share of voice and the most-cited domains and pages.
• Five Copilot modes: smart, chat, reasoning, study, search, plus Bing AI Mode (Copilot Search on Bing) with its full read-source list.

What is Microsoft Copilot?

Microsoft Copilot is Microsoft's conversational AI search assistant. Instead of a list of links it writes a detailed answer to your question and cites the web sources it drew from, and for shopping or local questions it adds product and place cards. This Actor returns that answer, its sources and its cards as structured data for any question you submit, so you can monitor, compare and report on how Copilot answers at scale.

Data you can extract

Each question returns one dataset row.

Field	Type	Description
query	String	The question you submitted
mode	String	The Copilot mode used: smart, chat, reasoning, study, search or bingAiMode
country	String	The country the question was asked from
status	String	success (complete) or partial (answer cut short)
responseComplete	Boolean	Whether Copilot finished the answer
responseMode	String	The mode Copilot actually answered in
model	String	The model behind the answer, when reported (Bing AI Mode)
title	String	The conversation title Copilot gave the answer
text	String	The full answer in Markdown
rawResponse	String	The same answer with [n] citation markers inline
reasoning	String	Copilot's reasoning steps, when it shows them
searchQueries	Array	The web searches Copilot ran to answer
tables	Array	Comparison tables, each with headers and rows
sources	Array	Cited sources: id, url, title, domain, section (citations or product), snippet
inlineCitations	Array	Each inline citation: id, url, title, textAnchor (the words it sits on)
researchSources	Array	Sources Copilot read but did not cite inline: url, title, domain, snippet
products	Array	Product cards: name, url, imageUrl, priceText, price, currency, seller, brand, rating, ratingCount, pros, cons, specifications, attributions
places	Array	Place cards: name, address, latitude, longitude, phone, category, rating, reviewCount, openState
recommendations	Array	Named picks and section headings for quick scanning
sections	Array	The answer split into heading and text sections
links	Array	Other links in the answer that are not cited sources
otherCards	Array	Any other answer cards Copilot returned, with their type
brandMentions	Array	Per brand: mentioned, mentionCount, firstPosition, snippets, rank, cited
sourceDomains	Array	The unique domains cited in the answer
citationCount, inlineCitationCount, productCount, placeCount, wordCount	Number	Convenient counts
warnings	Array	Notes, for example when an answer is incomplete
collectedAt, durationMs, requestId	String/Number	When the answer was collected, how long it took, and a stable id for the question

Questions Copilot does not answer are written to a separate Failed questions record, not to the dataset, so your results stay clean.

Example output

A shortened, illustrative row for best noise cancelling headphones under 300 dollars in smart mode (answer text abbreviated):

{
  "query": "best noise cancelling headphones under 300 dollars",
  "mode": "smart",
  "country": "US",
  "status": "success",
  "responseComplete": true,
  "responseMode": "chat",
  "title": "Best Noise-Cancelling Headphones Under $300",
  "text": "The strongest overall pick is the Sony WH-1000XM4 or XM5, but the best value pick is the Soundcore Q30, dramatically cheaper while still delivering real ANC. ...",
  "searchQueries": ["best noise cancelling headphones under 300 dollars"],
  "tables": [
    {
      "index": 0,
      "headers": ["Model", "ANC", "Battery", "Typical Price", "Best For"],
      "rows": [
        ["Sony WH-1000XM4", "Excellent", "30 hrs", "$278", "Best overall"],
        ["Bose QC45", "Very good", "24 hrs", "$279", "Comfort"],
        ["Soundcore Q30", "Good", "50 hrs", "$70", "Best value"]
      ]
    }
  ],
  "sources": [
    {
      "id": 1,
      "url": "https://www.claritypick.com/best-noise-cancelling-headphones-under-300/",
      "title": "5 Best Noise Cancelling Headphones Under $300 (2026 Buyer's Guide)",
      "domain": "claritypick.com",
      "section": "citations"
    }
  ],
  "inlineCitations": [
    {
      "id": 1,
      "url": "https://www.claritypick.com/best-noise-cancelling-headphones-under-300/",
      "title": "5 Best Noise Cancelling Headphones Under $300 (2026 Buyer's Guide)",
      "textAnchor": "Industry-leading ANC (blocks ~94% low-frequency noise)"
    }
  ],
  "products": [
    {
      "name": "Soundcore by Anker Space One Wireless Noise Cancelling Headphones",
      "url": "https://www.walmart.com/ip/soundcore-Space-One/1493272962",
      "priceText": "$99.99",
      "price": 99.99,
      "seller": "Walmart",
      "brand": "Soundcore by Anker",
      "rating": 4.41,
      "ratingCount": 16028,
      "pros": ["Comfortable and plush ear cushions", "Long battery life"],
      "cons": ["Low volume"]
    }
  ],
  "brandMentions": [
    {"brand": "Sony", "mentioned": true, "mentionCount": 6, "firstPosition": 97, "rank": 1, "cited": null},
    {"brand": "Bose", "mentioned": true, "mentionCount": 3, "firstPosition": 665, "rank": 2, "cited": null}
  ],
  "sourceDomains": ["claritypick.com", "themplsegotist.com", "amazon.com", "walmart.com"],
  "citationCount": 30,
  "inlineCitationCount": 10,
  "productCount": 6,
  "wordCount": 383,
  "collectedAt": "2026-10-30T08:48:31Z",
  "durationMs": 13578,
  "requestId": "4ec5720cc9e692ed0656c17634235bbf49cc10130619457555575095ae42f0d8"
}

Copilot modes

Choose one mode per run.

Mode	Best for
Smart (default)	A balanced answer with web search, product and place cards. The best default for most questions.
Chat	A quick, conversational answer.
Reasoning	A deeper, step-by-step answer for complex questions.
Study	A detailed, educational answer.
Search	A search-focused answer with sources.
Bing AI Mode	Copilot's answer as shown on Bing, with its full list of the sources it read.

Input

Input	Required	Description
Questions	Yes	One question per line. Up to 1,000 per run.
Copilot mode	No	Which mode answers each question. Default smart.
Country	No	The country each question is asked from. Copilot tailors local answers to it. Default United States.
Answers per question	No	Ask the same question several times (1 to 10) to measure a share of mentions instead of a single snapshot. Each answer is its own row and its own charge.
Brands to track	No	Brand or competitor names to measure in each answer. Add other spellings and a website after the name, for example HubSpot, HubSpot CRM, hubspot.com.

How to use Microsoft Copilot Scraper

1. Click Try for free and open the Input tab.
2. Type your questions, one per line, in Questions.
3. Pick a Copilot mode and a Country if you want something other than Smart and United States.
4. Optionally add your brand and competitors in Brands to track.
5. Click Start, then open the Dataset tab to preview and export as JSON, CSV, Excel or Google Sheets.
6. Automate it with schedules, webhooks, or the Apify API and MCP server.

Use cases

• AI search and brand monitoring (GEO/AEO). Track whether Microsoft Copilot mentions and cites your brand for the questions your customers ask, and how you rank against competitors.
• Citation and source research. See which publishers and pages Copilot trusts on your topics, and build a target list for content and digital PR.
• Competitive analysis. Compare how Copilot answers the same questions about you and your rivals, with share of voice over repeated runs.
• Product and local research. Collect the product and place cards Copilot surfaces for shopping and local questions, with prices, ratings and reviews.
• Change tracking. Schedule repeated runs and compare timestamped datasets to watch how Copilot's answers and sources shift over time.

Frequently asked questions

Does this require a Microsoft login or a Copilot subscription?

No. You do not need a Microsoft account, a Copilot subscription, or any API key. An Apify account is all that is needed to run the Actor.

Can answers change between runs?

Yes. AI answers and the sources behind them vary with time, region and demand. Repeated observations are more useful than treating a single answer as fixed. Ask the same question several times with Answers per question to measure a share of mentions.

Will every question return products, places or tables?

No. These depend on the question and the answer. A shopping question returns product cards; a local question returns place cards; a comparison often returns a table. Empty arrays mean Copilot did not return those for that question.

What happens when a question has no answer?

It is written to a separate Failed questions record with the question and a short reason, never to the main dataset, and it is not charged.

How much does it cost?

Usage is charged per answer delivered to your dataset. Questions Copilot does not answer are not charged. See the Pricing section on this page for the current rate.

Is it legal to scrape Microsoft Copilot answers?

Scraping publicly available data is generally legal in most jurisdictions, and Copilot answers are publicly accessible to anyone who asks. This Actor collects the same information a regular user would see. Always review the applicable terms of service and local regulations for your use case.

Is this an official Microsoft product?

No. This is an independent Actor and is not affiliated with or endorsed by Microsoft.

Use with an AI agent

Microsoft Copilot Scraper returns Microsoft Copilot answers as structured data. Give it queries (an array of questions, one per entry, up to 1,000). Optional: mode (smart, chat, reasoning, study, search, bingAiMode; default smart), country (two-letter code, default US), answersPerQuery (1 to 10, default 1; each answer is a separate row and a separate charge), and brands (names to track, each optionally with aliases and a website: HubSpot, HubSpot CRM, hubspot.com). Each question returns one dataset row with text, rawResponse, tables, sources, inlineCitations, researchSources, products, places, recommendations, sections, brandMentions, and counts. Questions Copilot does not answer go to a separate FAILED_QUERIES key-value record and are not charged. A run-level SUMMARY record carries brand mention rate, share of voice and the most-cited domains and pages. Usage is charged per answer delivered. No login or API key is required.

If this Actor did not do what you needed, add a short note in the feedback input on your next run saying what was missing. The run proceeds normally and the note reaches the maintainers.