You convert a PDF and get nothing, an empty file, or a page of blanks. The converter is not broken. Your PDF contains no text.
The ten-second test
Open the PDF in any reader and try to select a sentence with your mouse.
Individual words highlight? There is a text layer. Any converter can read it, and if one gave you nothing, try another.
A rectangle covers the whole block, or nothing selects? It is an image. The page is a photograph, and the words in it are arrangements of dark pixels, exactly as the words in a holiday snap of a road sign are.
This single check saves trying five converters that were all going to fail, because none of them can extract text that is not there.
How PDFs end up like this
Scanned on a scanner or a photocopier. The obvious case. The device photographs the page.
Photographed on a phone and "converted to PDF". Very common, and the conversion wraps a photograph in a PDF container without adding any text.
Exported from a design tool with text outlined or flattened. The one that catches people out, because the file was never near a scanner. Outlining converts letters into shapes: they look identical and are no longer text. This is also why some resumes and brochures parse as completely empty.
Printed to PDF by a driver set to rasterise. Occasionally a print-to-PDF path turns the whole page into an image, usually on older drivers or when a "print as image" option is enabled.
What actually helps
Find the original. Before anything else, ask whether the document exists somewhere as a real file. A statement from your bank's website rather than a scan of the posted copy; the Word file the PDF was made from; a fresh export with text not outlined. This is the best outcome by a wide margin and is skipped surprisingly often.
OCR, if you cannot. Optical character recognition reads letters out of an image and produces text. It is a genuinely different kind of software from a converter, and worth understanding before you rely on it:
It is recognition, not extraction, the output is a best guess. On clean, straight, printed text at a decent resolution, modern OCR is very accurate. On a phone photo at an angle, on faint print, on handwriting, or on anything with a complex layout, it is considerably less so. Numbers are the dangerous case: a misread 3 for an 8 in a financial document looks perfectly plausible and is silently wrong.
So always proof-read OCR output against the original, particularly the figures. Being sceptical of it is not pessimism; it is how the technology works.
A mixed document
Documents are often part text and part image, a digital contract with a scanned signature page, or a report with a photographed appendix. Those pages produce nothing while the rest converts normally.
The converter here reports which pages had no text layer rather than silently skipping them, so you know exactly what is missing instead of discovering it later.
Avoiding it in future
When you make a PDF, use File → Save as PDF or Export rather than Print → Save as PDF, and from a design tool check that the export does not outline or rasterise the text.
Then run the ten-second test on the finished file. If you can select a sentence, everyone downstream can work with your document, and if you cannot, you have found out now rather than when someone else needed to.