How to chat with a scanned PDF

You scan a contract at the office, save it as a PDF, upload it to a chat tool, and ask a question. The answer: there is no text in this document. Yet you can see the words perfectly well on screen. What went wrong?

Nothing is wrong with your file. It is just a different kind of PDF from the one most tools expect. This guide explains the difference, how askPDF handles scanned pages on the Plus and Pro plans, and what you can do to get the best results.

Two kinds of PDF

A PDF made by exporting from a word processor, a browser or accounting software contains text. Every letter is stored as a character, which is why you can select it, search it and copy it.

A PDF made by a scanner or a phone camera is different. Each page is a picture. The shapes of the letters are there as pixels, but the file holds no characters at all. A person reads the picture instantly. A program that looks for text finds nothing, because there is nothing of that kind to find.

Many documents are a mix. A report might be exported text with one scanned appendix. Those mixed files are the ones that surprise people most.

A page counts as image-only when it has almost no text of its own. A scanned page that carries a line of real text, such as a typed header, is indexed from that text alone.

How to tell which one you have

Open the PDF and try to select a sentence with your mouse, or use the search shortcut to look for a word you can see on the page. If the words highlight and the search finds them, the page has text. If you can only drag a box over a blurry rectangle, or the search finds nothing, the page is an image.

How askPDF reads image-only pages

On Plus and Pro, askPDF recognises these pages automatically when you upload, so you can ask questions about a scanned document like any other. You do not need to switch anything on or run a separate tool. Pages that already have text are read as usual, and pages that are only pictures are read from the image, so their words become something you can ask about.

After the upload finishes, you use the document like any other. You ask a question, the answer cites the page it came from, and you click the chip to open the PDF at that page. For a scanned page, that matters even more than usual, so you can compare the answer with what the page really shows.

On the Free plan, pages that already have text are indexed, but image-only pages are not read. If a document has no text at all, the upload will tell you that it needs a plan that reads scanned pages.

Reading scanned pages takes longer than reading text, so a long scan can take a few minutes. While it is processing you can already open the document and look at it, and the chat page shows its progress.

How good will the result be?

Reading a picture of text is a harder job than reading text, and the quality of the picture decides a lot. Clean, straight, well-lit scans of printed text usually come out well. Faint copies, small print, heavy skew, creased paper, handwriting and unusual fonts are harder, and some words may be misread.

So treat answers from scanned pages with a little extra care. Click the page chip and look at the original every time the answer matters, such as for a name, an amount or a date. The citation is there for exactly this reason.

There are also practical limits: there is a cap on how many image-only pages are read in one PDF, and a daily allowance of scanned pages per account, which is far above normal personal use. If a very long scan runs into the cap, the remaining pages are left unread, so split the file if you need all of it, and the daily allowance resets at midnight UTC.

Tips for a better scan

A few minutes of care when you scan saves a lot of confusion later.

• Flatten the page. Press the paper flat and avoid shadows along the fold.
• Light it evenly. A phone photo with a glare or a dark corner is hard to read. Daylight from a window works well.
• Hold the camera square. Shoot from directly above, with the page filling the frame and the text level.
• Use a scanner app. Apps that crop, straighten and sharpen pages give better results than a plain photo, and they can combine pages into one PDF.
• Choose a sensible resolution. Sharp enough to read the smallest print comfortably, but there is no need for huge files.
• Scan one document per file. It makes the cited pages easier to follow.

When a scan is not the right input

If you can get the original digital file, use it. A PDF exported from the source has real text, so it reads faster and more faithfully than any scan. Keep scans for the documents that only exist on paper: signed contracts, old letters, forms, printed books and notes.

Next steps

The guide on how to chat with a scanned PDF shows the steps and example questions, and the use cases page lists the other kinds of document you can ask about. If you are new to the whole idea, start with how to chat with a PDF.

Paper is not going away, and with image-only pages readable, it no longer has to be a dead end for your questions.