Webpage Text Extractor (Readability)

Turn any article URL into clean main-content text and markdown — nav, ads, sidebars, and footers stripped. The reader your AI agent needs for RAG, summarization, and content pipelines. No API key, pay per page.

▶ Live on the Apify Store — run it instantly, or call it as an agent tool via Apify MCP.

Why

LLM agents waste tokens on boilerplate. This returns just the readable article — as portable markdown (absolute links/images) and plain text — plus word count, reading time, and a readability score.

What you get (per page)

• markdown — clean GitHub-flavored markdown of the main content
• text — plain readable text
• title, byline, publishedAt, lang, excerpt
• wordCount, readingTimeMin, fleschReadingEase

Input

{ "url": "https://example.com/some-article", "outputFormat": "both" }

or bulk:

{ "urls": ["https://a.com/post", "https://b.com/post"], "maxUrls": 25 }

Output

{
  "url": "https://example.com/some-article",
  "title": "How web scraping works",
  "byline": "Jane Doe",
  "lang": "en",
  "markdown": "# How web scraping works\n\nWeb scraping is ...",
  "text": "How web scraping works. Web scraping is ...",
  "wordCount": 1240,
  "readingTimeMin": 6,
  "fleschReadingEase": 58.2
}

Notes

Uses a readability heuristic (semantic containers + text-density scoring) — works on most articles and blogs without a headless browser, so it's fast and cheap. Returns only the public content of the URL you provide.

Pricing

$0.03 per page extracted — billed as the page-extracted event. Bulk runs are capped by maxUrls, which is also your budget cap.

FAQ

How do I convert an article URL to Markdown? Pass it in with outputFormat: "markdown" (or "both") — the markdown field is clean GitHub-flavored Markdown of just the main content, with links and images resolved to absolute URLs so it stays portable.

Why not just fetch the page HTML for my LLM? Raw HTML is mostly boilerplate — nav, ads, sidebars, footers — and LLM tokens cost money. Readability extraction sends the model only the article, typically a fraction of the raw page's size.

What metadata comes with the text? title, byline, publishedAt, lang, excerpt, plus wordCount, readingTimeMin, and a Flesch reading-ease score per page.

Does it work on JavaScript-rendered pages? It parses the served HTML without a browser — which covers most articles and blogs and keeps it fast and cheap. Pages that render their entire article client-side need a browser-based scraper instead.

Can I batch a reading list? Yes — pass an array in urls (capped by maxUrls), one record per article, ready for a RAG ingestion pipeline or a summarizer.