PDF to Markdown for AI tools
For ChatGPT, Claude or a retrieval pipeline, convert the PDF to Markdown first: ConvCC writes the text in reading order with headings, lists and tables, and removes running headers and page numbers. The model gets the document's structure, with fewer tokens and no text from two columns mixed together.
2 files an hour free, short or long. No account needed.
Updated:
Private files. Never used for training. Deleted automatically, or right away when you press Delete now.
What you get
- Headings to split the document into sections (chunks) that keep their meaning.
- Tables as Markdown tables, which models read far better than columns of loose numbers.
- No repeated page furniture: every page header you remove is tokens you don't pay for.
- Need the page of each block for citations? Use PDF to JSON instead.
Why raw PDF text confuses a model
Copying text out of a PDF gives you lines in the order they were drawn. On a two-column page that can be half a line from each column; on a report every page repeats the same header. A model reads all of it as the document, and answers from it.
Start free. Go Pro for scans and no hourly limit.
FAQ
Is it free?
Yes: 2 files an hour with any tool, short or long, each up to 200 MB, with no account. Pro removes the hourly limit and adds OCR for scanned PDFs.
Does it work with scanned PDFs?
Yes, on Pro: pages that are only a picture are read with OCR (English, Spanish, French, German, Portuguese, Italian and Dutch), and the report lists them so you can check names and numbers. On Free, a scanned page has no text to export.
What happens to my file?
Your file is deleted as soon as it is converted, or 2 hours after its last conversion. The result is deleted 1 hour after your first download, or 24 hours after it's ready. Files are never published, and never used for training.