How to count tokens in a PDF

Models count the extracted text of a PDF, not the file itself, so the token count depends on how much text the document holds once it is parsed.

The quick method

1. Open the PDF and select all text (Ctrl/Cmd + A), then copy it.
2. Paste it into the token counter for an exact GPT count and estimates for other models.
3. For scanned PDFs with no text layer, run OCR first, since images carry no extractable text.

Rule of thumb

A dense single-spaced page is roughly 500 to 700 tokens. A 20-page report is therefore often 10,000 to 14,000 tokens, enough to matter against smaller context windows. Convert precisely on the tokens per page tool.

Direct in-browser PDF upload is planned. Until then, paste the extracted text, which also keeps your document on your own device.