PDF to Markdown: Tables, OCR & RAG Chunks

What does PDF to Markdown do?

PDF to Markdown converts PDF files from URLs into clean, structured Markdown or plain text. It keeps what matters for reading and for AI: headings, paragraphs in the right reading order (including two-column layouts), bulleted and numbered lists, and tables as Markdown tables. Scanned pages are read with OCR in 12 languages. You get one result per PDF, per page, or as ready-to-embed chunks for RAG.

Run it from Apify Console, call it through the Apify API, schedule it, or plug it into LangChain, LlamaIndex, Make, Zapier and other integrations.

Why use PDF to Markdown?

• RAG and AI assistants: feed LLMs clean text instead of broken PDF extracts. Chunk mode splits on paragraph boundaries with overlap.
• Research and analysis: turn reports, papers and filings into searchable text with headings preserved.
• Data pipelines: pull tables out as Markdown, or use plain text with tab-separated table rows.
• Scanned archives: OCR turns image-only pages into text automatically.

How to use PDF to Markdown

1. Click Try for free.
2. Add your PDF links under PDF URLs, or upload a text file with one URL per line.
3. Pick Markdown or Plain text, and choose one result per document, page or chunk.
4. Click Start. Download the results as JSON, CSV, Excel or HTML, or read them through the API.

Input

{
  "pdfUrls": [{ "url": "https://arxiv.org/pdf/1706.03762" }],
  "outputFormat": "markdown",
  "splitBy": "chunk",
  "chunkSize": 2000,
  "chunkOverlap": 200,
  "extractTables": true,
  "ocr": "auto",
  "ocrLanguages": ["eng"],
  "pages": "1-10",
  "saveMarkdownFiles": false
}

See the Input tab for all options, including a page selection (e.g. 1-5, 8), a page limit per PDF, and a maximum file size.

Output

You can download the dataset in various formats such as JSON, HTML, CSV, or Excel. Example of one chunk:

{
  "url": "https://arxiv.org/pdf/1706.03762",
  "fileName": "1706.03762.pdf",
  "status": "ok",
  "pageCount": 15,
  "pagesProcessed": 15,
  "ocrPages": 0,
  "tablesFound": 6,
  "metadata": { "title": null, "creator": "LaTeX with hyperref", "producer": "pdfTeX-1.40.25", "createdAt": "2024-04-10T21:11:43" },
  "chunkIndex": 0,
  "chunkCount": 28,
  "content": "Provided proper attribution is provided, Google hereby grants permission to reproduce the tables and figures in this paper solely for use in journalistic or scholarly works.\n\n# Attention Is All You Need\n\n…",
  "charCount": 1756,
  "processedAt": "2026-09-17T10:12:03+00:00"
}

With Also save each document as a file on, every PDF's full Markdown is also stored as a .md file, and fileUrl links to it.

Data fields

Field	Description
url, fileName	Source PDF
status, error	ok, or error with the reason (e.g. not a PDF, too large)
pageCount, pagesProcessed	Pages in the PDF and pages converted
ocrPages, tablesFound	How many pages needed OCR and how many tables were extracted
metadata	Title, author, subject, creator, producer, created and modified dates
content	The Markdown or text for the document, page or chunk
pageNumber	In page mode
chunkIndex, chunkCount	In chunk mode
wordCount, charCount	Size of the content
fileUrl	Link to the saved .md/.txt file, when enabled

How much does it cost?

Pay-per-event pricing, so you pay only for what you convert:

• $2.00 per 1,000 PDFs
• $0.05 per 1,000 pages
• $2.00 per 1,000 OCR pages (only pages that needed OCR)

For example, 1,000 ten-page digital PDFs cost about $2.50. A 100-page scanned document costs about $0.21. Apify's free plan includes monthly usage credits to try it. Set a maximum cost per run in the run options and the Actor stops cleanly when it's reached.

Tips

• Use Pages to convert only the parts you need, such as the first 10 pages of long reports.
• Choose only the OCR languages actually in your scans; fewer languages give better accuracy.
• Set OCR to Off for digital PDFs when you're sure there are no scanned pages, to guarantee no OCR charges.
• For RAG, 1,000–2,000 character chunks with 10% overlap are a good starting point.

FAQ, limitations and support

What kinds of tables are extracted? Tables drawn with lines or cell borders are extracted as tables. Tables laid out only with spacing (common in academic papers) come through as text instead.

How good is the reading order? Single- and two-column layouts, like reports and academic papers, come out in reading order. Very complex magazine-style layouts, text inside figures, and equations may come out in a less natural order.

Can it read password-protected PDFs? No. Encrypted PDFs that need a password return an error.

Where do the PDFs come from? Only from the URLs you provide. You're responsible for having the right to process those documents.

Found a PDF that converts badly? Open an issue in the Issues tab with the link and it will be looked at promptly.