About PDF.chat

PDF.chat lets you upload any PDF or document and chat with it — ask questions and get instant answers, each cited to the exact page, in over 100 languages.

Finding the answer buried in a long document is slow, and most AI tools answer from memory and make things up. We built PDF.chat to fix that: every answer is grounded in your document and cites the page it came from, so you can click any citation and verify it yourself. A fast engine handles everyday questions, and a premium AI engine tackles the hardest documents.

No signup to try, generous free limits, multi-document chat, the original shown beside your conversation, and a simple API for developers.

How PDF.chat works

Your conversations are powered by the leading AI models — Claude, GPT, Gemini and more — and on Pro you can compare them side by side on the same question. To read your documents we build on the best open source: digital PDFs are parsed instantly, and scanned pages are read by OCR models we run on our own infrastructure. On top we add the layer that makes it trustworthy — retrieval over long documents, page-level citations, and side-by-side reading.

  • PaddleOCR (PP-OCRv6) Apache-2.0
    Reads scanned and photographed pages so you can chat with paper documents too. Excellent on printed text in 100+ languages, with per-language models for Latin, CJK, Arabic, Cyrillic and Indic scripts.
  • PaddleOCR-VL & DeepSeek-OCR open source
    Open vision-language models that read handwriting, math and complex layouts, so even tricky scanned pages become chattable. We run these OCR models on our own infrastructure.
  • PyMuPDF open source
    Reads the text layer of digital PDFs instantly and losslessly, so clean PDFs are ready to chat with immediately, and renders scanned pages for the OCR models.
  • python-docx & openpyxl open source
    Read Word documents directly (no OCR needed) so you can chat with them, and write real Excel spreadsheets when you export tables.

Standing on open source means you get state-of-the-art accuracy that keeps improving with the community, not a black box. Thank you to everyone who builds these projects.

Frequently asked questions

Yes. You can upload a PDF and start chatting with no signup at all. Create a free account for more questions each month, or upgrade for larger documents, multi-document chat, the highest-accuracy AI models, and API access.

PDFs of every kind, plus Word, PowerPoint, text files, and images (PNG, JPG, WEBP, GIF, BMP, TIFF). You can chat with documents in over 100 languages, including Latin, Chinese, Japanese, Korean, Arabic, Cyrillic, and Indic scripts, and ask your questions in your own language. Scanned PDFs are read with OCR first.

Yes. PDF.chat understands tables and equations in your document, so you can ask about figures, totals, and formulas and get cited answers. You can also export tables as Markdown or CSV and equations as LaTeX.

Plain text, Markdown, Word (DOCX), searchable PDF, CSV, and JSON, all from the result screen, plus one-click copy to clipboard.

Your files are processed only to answer your questions and deleted automatically. We never sell or share your documents, and we don't train on them.

Yes. Our premium AI engine reads handwritten notes, forms, and mixed printed & handwritten pages, so you can chat with them too. Clear handwriting works best, and every answer cites the page so you can check it against your original.

Yes. PDF.chat indexes long documents and retrieves the most relevant pages for each question, so it works on hundred-page reports, textbooks and dockets. Pages with a text layer are read instantly and only scanned pages go through OCR, and larger documents process in the background while you wait.

There is a genuinely free tier you can use without signing up. Paid plans start at $5/month and add more pages, multi-document chat, the highest-accuracy AI models, and API access. Usage is metered in pages, where one page of a document equals one page.

Many AI tools answer from memory and make things up. PDF.chat grounds every answer in your document and cites the exact page, so you can verify it. It pairs a fast engine for everyday questions with a premium AI engine for the hardest documents, shows the original beside your conversation, and reads scanned PDFs with OCR so you can chat with them too.

Yes. The workspace shows your original document side by side with the chat. Every answer cites the page it came from — click a citation and the page scrolls into view with the passage highlighted, so you can check it yourself.

No. PDF.chat runs entirely in your browser, just upload a PDF and start chatting. There is nothing to download or install.

Yes. A simple REST API lets you upload a document and chat with it programmatically, getting back answers with page citations. You can also process documents into text, Markdown, tables, and JSON. See the API docs for details.

Yes. The premium AI engine reads charts, diagrams, and images inside your PDF, so you can ask what a figure shows, summarize a chart, or pull numbers out of a graph — all cited to the page.
Try it now Contact Us
Rate this page
★ ★ ★ ★ ★
5.0/5 (0)

What could we improve? Your feedback helps us fix issues.