Visible words can be pixels rather than PDF text
A PDF can contain text objects, page-sized images, or a mixture of both. A scanned letter may look identical to a document exported from a word processor while storing its words only as image pixels. Text extraction reads the stored characters; it does not interpret the shapes of letters in a photograph.
Try selecting and copying a known sentence in a PDF reader as a quick initial check. Failure to select it suggests that the page may be image-only, but it is not a complete diagnosis: text encoding, permissions and reader behaviour can also matter. Check more than the first page, since a document can mix scanned and text pages.
| Symptom | Possible cause | Next check |
|---|---|---|
| Empty output for a visible scanned page | No extractable text layer | Use OCR on a copy and inspect the recognised words |
| Some pages work and others are empty | Mixed text and image-only pages | Check each page and the selected range |
| Words appear in an odd order | Columns or positioned text objects | Compare the TXT against the visual page |
| The file cannot be opened | Protected or unreadable PDF | Use an authorised readable copy; do not treat this as empty text |
Check page selection before changing the document
This tool reads selected pages in document order, not the order in which page numbers were typed. Plain text does not preserve PDF columns, tables, margins or interactive fields. A successful extraction can still need editing before it becomes a useful transcript.
- Open PDF to Text and choose the PDF. Wait for the initial extraction to finish.
- Leave Pages empty to read the whole document, or enter the intended range, such as 1-3, 5. Confirm that those pages actually contain the text you need.
- Enable page-start marks when you need to identify where each page's output begins.
- Compare a few known sentences, names and numbers with the original pages. Copy the text or download the TXT only after that comparison.
A public example with and without a text layer
Our four-page public PDF contains the visible test phrase SECRET_SAMPLE and other text. The saved all-page TXT sample contains 161 characters. A separate public page-image rebuild of the same source has no extractable text items, even though its rendered pages still show the words.
These are authored examples, not user documents. The image-only PDF illustrates why visible writing is insufficient for direct extraction. Neither an empty result nor a successful search test establishes whether a document has been fully sanitised; this article concerns retrieving text, not certifying hidden-data removal.
| Sample | Pages | Text extraction |
|---|---|---|
| Original PDF | 4 | Text present; public TXT has 161 characters |
| Page-image rebuild | 4 | No text items |
When OCR is the next step
OCR recognises letters from image pixels and can add a searchable text layer or produce a transcript. PDF to Text on this site does not include that recognition step. If the selected page is image-only, run OCR in a tool that explicitly provides it, using the correct document language.
Check recognised dates, totals, names, punctuation and table rows against the scan. Low resolution, rotation, handwriting and unusual layouts can produce errors. Keep the original scan and treat the OCR result as a derived copy that needs review, especially when a number will be used elsewhere.
Avoid removing the text layer before extracting it
Use the original readable PDF when you have it. Rendering pages to JPG, putting those JPGs back into a PDF, or using a page-image compression path can remove the original text layer. A new PDF extension does not recreate the characters that were discarded.
If one page extracts incorrectly while others work, isolate that page for diagnosis and compare another reader or the originating document. If the file cannot be opened at all, resolve that access or file-integrity issue first. An opening error and an empty but readable page are different outcomes.