Skip to content

Extract API

Extract uses computer vision and natural language processing to read a web page the way a person would, then return its contents as clean, structured JSON — no rules or per-site configuration required.

Every Extract endpoint is a single GET request: pass your token and the url you want to read, and Diffbot renders the page, classifies it, and returns the structured objects it finds. Extract has a base cost of one credit per request (two if a proxy is used). See how credits work.

Automatic API

If you aren't sure what type of content is at a URL, start with Analyze — it classifies the page with machine learning and routes it to the right page-type extractor automatically.

  • Analyze — classify any page and extract it as the appropriate type.

Page-type APIs

If you already know the content type (or want to force extraction as a specific type), call a page-type endpoint directly for a guaranteed extraction.

  • Article — news articles, blog posts, and other text-heavy pages: author, date, sentiment, tags, images, and clean body text.
  • Product — e-commerce product pages: price, availability, specs, images, brand, and reviews.
  • Image — the primary image(s) on a page, with dimensions, download URLs, and recognition tags.
  • Video — video pages: metadata, thumbnails, direct video URL, embed code, and view count.
  • Discussion — threads of posts: forum threads, comment sections, and product reviews.
  • Event (Beta) — single-day online and in-person events: dates, location, and description.
  • List (Beta) — listings pages: news indexes, product listings, and search-results pages.
  • Job (Beta) — job postings: employer, location, requirements, skills, and tasks.

Authentication

Authenticate every request with your Diffbot token, passed as the token query parameter. See Authentication for where to find your token.