Practical PDF tasks
PDF text extraction came back blank? Check for a text layer first
Check text page by page, export TXT or Markdown, and recognize when scanned pages need a separate OCR step.
An eighteen-page PDF lets you copy text from the first twelve pages, but the last six give you nothing. Select All, Copy and Paste produces half a document, which can look like a successful extraction until somebody notices the missing ending. The first question is whether each page has a text layer at all.
Test one good page and one suspect page
Open Extract Text from PDF and load the file. Run extraction on page 1 and compare the result with its visible heading and paragraphs. Then select a page that appears to be a scan. Check its status and character count. The tool reports results by page; success on the first page says nothing about the rest of the file.
Visible letters in a PDF do not necessarily exist as selectable text. A scanned page may be one large image. Text extraction cannot identify its letters on its own. Turning image pixels into words requires OCR, or optical character recognition; this extraction tool is not an OCR tool.
Export according to what you need next
For a quotation, inspect and copy a single page. To hand the available text to a colleague, export TXT. Markdown retains page headings that make it easier to trace passages back to the source. JSON is useful if another program needs page-level results. Open the downloaded file and check for missing page boundaries, table columns that ran together, and headers or footers mixed into the body.
Multi-column pages need special attention. What reads left column then right column to a person may be stored in another order inside the PDF. Try the readable and layout modes, and the reading-order options, on one representative page before processing the full file. None of these settings guarantees that a complex table or footnotes become clean structured data.
What to do with scan-only pages
If the last six pages have no text layer, look for the source document or an already OCR-processed copy. Otherwise, use a separate OCR workflow on those pages. Review names, amounts, dates and identifiers after OCR. Characters such as 0/O and 1/l are easy to confuse; a long output is not proof of accuracy.
For this example, the handoff note should say: pages 1–12 were extracted, pages 13–18 still need OCR. That is more useful than presenting a partial TXT file as though it contains the whole PDF.