How to extract text from a PDF — the four routes that actually work
Updated 13 August 2026
“Extracting text” is four different jobs. Running text: convert the PDF to Word, then save as plain text. Tables: export to Excel instead. A scan: run OCR first — it can also download the recognised text. One paragraph: open the file and select and copy. Every route runs in your browser.
Get text out of a PDF four ways: convert to Word, export tables to Excel, OCR a scan, or select and copy. Free, in your browser, no upload.
UnboundPDF is a free suite of 43 PDF and image tools that run entirely in your browser — merge, split, compress, edit text, OCR in 126 languages, redact, sign, convert and archive to PDF/A. Your document is read and written by the page on your own device; there is no document-upload endpoint in the core tools, no account, no watermark and no daily cap. Every result can be checked — with the Network tab, or with the Document Passport the Workspace writes for a chain of steps.
- Decide which job you have: running text, a table, a scanned page, or a single paragraph.
- Running text → PDF to Word: export a .docx, then use Save As → plain text in Word, Pages or Google Docs.
- Tables → PDF to Excel: it detects the column boundaries per page and exports a standard .xlsx workbook (CSV also available).
- Scanned pages → OCR PDF first: recognition runs on your device, and you can download the recognised text or the searchable PDF.
There is no single “PDF to text” button here — because “extract the text” is four different jobs with four different right answers. This page names each and sends you to the tool that does it, in your browser.
Four jobs that all sound like “extract the text”
People typing extract text from PDF want different things: the wording of a report, the numbers in a table, the words trapped inside a scan, or one paragraph for an email. A tool that pretends these are one job does three of them badly. Pick your row.
| What you actually want | Route | What you get |
|---|---|---|
| Running text (report, letter, article) | PDF to Word | An editable .docx — save it as plain text from there |
| A table of figures or rows | PDF to Excel | A standard .xlsx workbook with columns kept as columns (CSV also available) |
| A scanned or photographed page | OCR PDF | A searchable PDF, plus the recognised text |
| One paragraph, right now | Your viewer or the PDF Editor | Selected text on your clipboard |
| The figures, not the words | Extract Images | The embedded pictures as separate files |
Running text: convert to Word, then save as plain text
For anything paragraph-shaped, PDF to Word is the route: text flows out as real paragraphs, headings stay bold and larger, and embedded JPEG photos ride along at their original quality. From that .docx, every word processor — Word, Pages, Google Docs — offers Save As → plain text, the step people expect a “PDF to text” button to do for them. One extra click, and the structure survives long enough to show you which lines were headings before you flatten them.
Tables: the Excel route, not the Word route
A table pushed through a text converter comes out as a wall of words in reading order — technically the text, practically useless. PDF to Excel does the other thing: it detects column boundaries per page from where the rows agree, then exports a standard .xlsx workbook, with numbers and dates stored as values, not text (CSV is also available). Prose paragraphs are kept whole rather than forced into columns. Bank statement, invoice register, results table — start here. Why typed cells (not text pretending to be numbers) matter for the formulas you run next: typed cells, not just text.
Scanned PDFs: there is nothing to extract yet
A scan holds pixels shaped like letters, not letters, which is why PDF to Word returns an empty document instead of guessing. OCR PDF is the missing first step: a vendored copy of Tesseract runs inside your browser, reads the pages in any of 126 languages, and lays an invisible selectable layer over the originals. Download the searchable PDF and copy from it, or take the recognised text straight out. Full walkthrough: OCR a scanned PDF. To change the words rather than lift them, see editing scanned PDF text; crooked pages read better after Deskew.
One paragraph: select and copy
If the file has a real text layer and you need a quote rather than a corpus, open it in your browser’s viewer or the PDF Editor, drag across the words, copy. No conversion, no fidelity loss. If dragging selects nothing you have a scan, or a protected file — those need Unlock PDF and the password. And if what you want is a figure, not a sentence, use Extract Images — embedded images vs PDF to JPG covers which.
Extraction sees more than the page shows
A text layer can hold words the page does not display — text under a drawn box, or an earlier wording left by a sloppy edit. Same mechanism as can deleted PDF text be recovered?, and it cuts both ways. It is also why extracting on your own device matters: these routes work in the browser tab rather than posting your file to a server, which you can check with the Network-tab method in the PDF Privacy Lab.
Scope
OCR quality follows scan quality — clean, high-resolution pages in the selected language read well; blurry or skewed photos produce more errors; handwriting is not recognised. Column detection reads text geometry, so merged cells and heavily styled tables may need tidying afterwards. A .docx carries the content; multi-column pages re-flow and vector graphics are not carried across — that is the difference between the two document models. More in Guides.
Frequently asked questions
Does UnboundPDF have a PDF-to-text tool?
Not a single dedicated .txt tool. The routes are PDF to Word for running text (save the .docx as plain text from any word processor), PDF to Excel for tables, and OCR PDF for scans — OCR can download the recognised text directly.
Why can’t I select or copy the text in my PDF?
Almost always because the page is a scan: pixels shaped like letters, not letters. Run OCR PDF to add a selectable text layer, then copy normally. A password-protected file is the other common reason — unlock it first.
Will the extracted text keep the original layout?
No. PDF to Word recovers the content, not the pixel layout: paragraphs re-flow as normal Word paragraphs, headings stay bold and larger, embedded JPEG photos ride along. Columns and vector graphics are not reproduced.
Is the PDF uploaded to extract its text?
No. The conversion and OCR code runs in your browser tab; the document is not uploaded to an UnboundPDF processing server. OCR does fetch its recognition engine (about 14 MB) the first time — program code, not your file, and visible in the Network tab.