How to Extract Text from a PDF (2026)

Two Types of PDFs, Two Extraction Paths

There are two fundamentally different kinds of PDFs when it comes to text extraction. A text-based PDF stores text as characters with positions, fonts, and styles — the same way a Word document does internally. Extracting from this type is fast, lossless, and produces clean output because the text is already there, just wrapped in PDF structure.

A scanned PDF is the opposite: it stores text as images with no underlying character data. Extracting text requires OCR — running the images through a recognition engine that reads pixels and outputs characters. This is slower, imperfect (98–99% accuracy on clean scans, worse on poor ones), and always requires a separate processing step.

The quick test: open your PDF, select a line of text with your cursor, and try to copy it. If clean text appears in the paste, you have a text-based PDF — use direct extraction. If nothing highlights or you get garbage, you have a scanned PDF — use OCR first via the PDF OCR tool, then extract from the OCR'd version.

Method 1: UDT PDF Text Extractor (Free, Browser-Based)

The UDT PDF Text Extractor pulls text from any text-based PDF directly in your browser using pdf.js. Upload (locally — nothing leaves your browser), and the tool outputs plain text ready to paste anywhere: email, Word, a terminal, another document. You can choose between preserving the original formatting (line breaks where they appear visually) or collapsing to natural paragraphs (useful for feeding into other processing).

The workflow: drop your PDF into the tool, choose output format (plain text, JSON with positional data, or Markdown), and download or copy. For scanned PDFs, the tool will detect the lack of a text layer and route you to OCR first — you can't extract text that isn't there.

One quality-of-life feature worth mentioning: the tool handles column detection automatically for most layouts. A two-column article extracts with the left column fully read before the right column starts, rather than producing the classic "left-right-left-right" zigzag you get from naive extractors.

Method 2: Acrobat and Other Alternatives

Acrobat Reader (free) can export PDF text via File → Export To → Text. Acrobat Pro adds Save As → More Options → Text (Plain) with additional formatting controls. Both handle text-based PDFs well. For scanned PDFs, Acrobat Pro runs OCR automatically during export; Reader doesn't.

Microsoft Word (2013+) can open PDFs directly, treating them as editable documents. Word's conversion is often the best-in-class for preserving formatting (fonts, tables, images), though simple text extraction is overkill if you just need the words. Google Docs has similar PDF import capabilities.

Command-line tools like pdftotext (from Poppler) are the workhorses for automated extraction. pdftotext handles layout preservation well, supports columns, and is what most document-processing pipelines use under the hood. Install via Homebrew on macOS, apt on Linux, or prebuilt binaries on Windows.

Handling Tables, Columns, and Layout

Tables are the hardest part of text extraction. A PDF table is just text positioned in a grid — there's no underlying "this is a table" structure the way there is in HTML or Word. Extraction tools guess at table structure by looking for aligned text runs, which works for clean tables and fails for complex ones with merged cells, nested columns, or inconsistent spacing.

For critical table extraction (financial data, scientific tables, anything where getting the structure right matters), specialized tools like Tabula or Camelot outperform general text extractors. They detect table boundaries, infer columns from x-coordinate clustering, and output CSV or JSON directly. General extraction tools give you text in reading order; table-specific tools give you rows and columns.

Multi-column layouts (magazines, academic papers, brochures) need column-aware extraction. A naive tool reads left-to-right across all columns on each row, producing nonsense. A better tool detects column boundaries and reads each column top-to-bottom in sequence. If your extracted text looks like interleaved fragments, you're probably using a tool that doesn't do column detection — switch to one that does.

Common Pitfalls and Cleanups

Ligatures are the most common complaint. PDFs often render "fi" and "fl" as single glyphs to look better typographically, and some extractors output these as question marks, boxes, or nothing at all. Modern extractors (including pdf.js-based tools) handle ligatures correctly; older ones or command-line defaults may not. If your output has missing letter-pairs, check ligature handling in your tool's settings.

Hyphenated line breaks are another trap. When a word wraps across lines as "how- ever" at the end of one line and start of the next, a careless extractor keeps the hyphen and newline in the output, producing "how- ever" in your text. Good extractors rejoin these automatically; if yours doesn't, a simple find-and-replace for "- " fixes most cases.

Header and footer repetition is the third common issue. Page numbers, running headers, and footers appear on every page, and naive extraction includes them in the output — producing "Page 1\nThe quick brown fox jumped...\nPage 2\n...over the lazy dog." Either skip headers and footers in the tool settings or strip them with a regex after extraction.