A PDF text extractor pulls the words out of a PDF so you can copy, search, or reuse them. Which tool works depends on one thing: whether the PDF has a real text layer or is a scanned image of a page. With a text layer, almost any method works, including copy and paste. A scan needs OCR (optical character recognition) to read the characters first. To check in 30 seconds, open the PDF and try to select one word. If the word highlights, the text is there. If you can only select the whole page as a picture, or nothing at all, it is a scan.
super.AI publishes this guide and builds document processing software. Tool details come from each vendor's or project's own documentation, checked October 15, 2026, and may change. We do not rank these tools or make accuracy claims about them. All product names are trademarks of their owners.
#Native PDF or scanned PDF: check this first
A native PDF was created by software, such as a Word export or an accounting system, and stores characters with their positions. A scanned PDF holds a picture of each page, so there is nothing to extract until optical character recognition (OCR) reads the image.
- Select a word. If one word highlights on its own, the PDF has a text layer.
- Search for a word you can see. If Ctrl+F (Cmd+F on a Mac) finds nothing, the text is probably an image.
- Paste what you copied. Garbled text or odd symbols suggest a damaged text layer, and OCR may do better.
One catch: some scans already carry a hidden OCR layer, for example from Adobe's online OCR tool, which creates "a searchable PDF." Check numbers and names in that text against the page image.
- Copy and paste. Works on native PDFs in any viewer. Limit: multi-column pages and tables often paste out of order.
- Adobe Acrobat. Adobe's help center explains how to "convert PDF files to plain text or XML format" in Acrobat. Limit: for scans, Adobe documents a separate step to recognize text in scanned documents.
- Adobe's online OCR tool. Adobe says it lets you "recognize text in a PDF document for free," producing a searchable PDF. Limit: you sign in to download the result.
- Google Drive and Google Docs. Right-click a PDF in Drive and choose Open with, then Google Docs. Drive converts it to text, scans included. Limit: Google says the file "should be 2 MB or smaller," and tables and columns are not likely to be detected.
- Microsoft Word. Open the PDF with File > Open, and Word converts it. Limit: Microsoft says this "works best with PDFs that are mostly text."
- Preview and Live Text on a Mac. In Preview, choose Tools > Text Selection, drag over the text, and choose Edit > Copy. For text in images, Live Text lets you select and copy text in Preview. Limit: Apple says "Live Text isn't available in all regions or languages."
- Free online converters. Many sites convert PDFs to text at no cost, and some add OCR. Limit: your file goes to someone else's server, so for contracts or personal data, read the privacy policy first or stay offline.
For extracting text from PDFs in bulk, a script beats clicking.
- pdftotext (Poppler). A command-line tool that "converts Portable Document Format (PDF) files to plain text," by default "in reading order." The -layout option tries to keep "the original physical layout of the text," as in pdftotext -layout invoice.pdf invoice.txt.
- pypdf. A Python library: call extract_text() on each page. Its docs say it "will also never be able to extract text from images" and point to OCR software such as Tesseract.
- pdfplumber. A Python library that gives "detailed information about each text character, rectangle, and line," plus table extraction. It "works best on machine-generated, rather than scanned, PDFs."
- PyMuPDF. A Python library whose get_text() returns plain text, or blocks and words "with position information." Its OCR support relies on Tesseract, which "must be installed as a separate application."
- Apache PDFBox. "An open source Java tool for working with PDF documents" that can "extract Unicode text from PDF files," with command-line utilities.
- Apache Tika. A toolkit that "detects and extracts metadata and text from over a thousand different file types." It suits folders that mix PDFs and Office files.
- Tesseract. An OCR engine that reads images "including PNG, JPEG and TIFF" and recognizes "more than 100 languages" out of the box. It does not open PDFs directly, so render each page to an image first.
Cloud APIs run OCR at scale and return text with its position on the page, as JSON.
- Amazon Textract (DetectDocumentText). AWS says Textract returns "the lines and words of detected text" and their location on the page. Each block includes confidence and geometry, and Textract marks text as handwritten or printed. Synchronous calls take one-page PDFs, and the asynchronous StartDocumentTextDetection operation handles up to 3,000 pages. AWS's pricing page lists $0.0015 per page for the first one million pages in US West (Oregon). New AWS customers get a Free Tier of 1,000 pages a month for three months.
- Google Cloud Document AI (Enterprise Document OCR). Google says it "can detect blocks, paragraphs, lines, words, and symbols from PDFs and images." For digital PDFs it can "extract embedded text" instead of running OCR. Google's pricing page lists $1.50 per 1,000 pages for up to 5,000,000 pages a month.
- Azure Document Intelligence in Foundry Tools (Read model). Microsoft says the Read model "extracts print and handwritten text from PDF documents and scanned images" and also reads Word, Excel, PowerPoint, and HTML files. It returns lines and words with "bounding polygon coordinates" and confidence. Microsoft's pricing page lists up to 500 free pages a month on its free tier.
A PDF is built to look right, not to be read by software. The pypdf docs say "The text within a PDF document is absolutely positioned," with "no information what the header, footer, page numbers, tables, and paragraphs are." Common problems and fixes:
- Wrong reading order. An extractor may read straight across two columns and mix their lines. PyMuPDF's docs note that text "may not appear in any particular reading order" and offer a sort option. Try both a reading-order mode and a layout mode.
- Tables flattened into lines. Rows become runs of text, and wrapped cells split in two. Use a tool with table extraction, such as pdfplumber, or see our post on automating table extraction from PDFs and scanned images.
- Repeated headers and footers. Strip lines that repeat on most pages, or crop pages to the body area.
- Hyphenated words. "Extrac-" and "tion" need rejoining. The pdftotext default mode tries to undo hyphenation, per its documentation.
- Ligatures. Some fonts draw "fi" or "fl" as one character, so "office" may come out looking wrong. Unicode NFKC normalization splits these back into separate letters.
- Rotated pages and poor scans. OCR struggles with sideways, skewed, or blurry pages. Google says "Documents must be right-side up," and Tesseract's README notes you may need to "improve the quality of the image." Rotate, straighten, or rescan before OCR.
#When you need data, not text
Often the goal is not the text. You want the invoice number, total, and line items in a spreadsheet, or form fields in your system. That is the step from "extract text from PDF" to "extract data from PDF," and raw text is only the start.
There are two common ways to turn text into fields:
- By position (templates). You define where each field sits on a known layout. It works for one fixed form but breaks when a vendor moves a field, and each new layout needs its own template.
- By meaning. The software finds each value from labels and context, such as the amount next to "Total due," wherever it appears. This copes better with layout variety, but no method is right every time.
So plan for review. Textract and the Azure Read model return a confidence score for each word. The same idea works for fields: send uncertain values to a person and pass the rest through. For how this breaks in practice, read why invoice and PDF data extraction fails. Our earlier guide to extracting data from PDFs covers the basics.
super.AI Flows is built for this step. AI extraction finds values by their meaning rather than their position on the page, so you do not need a separate template for each layout. Low-confidence results go to human review, and corrections carry into the next run. Read more about data extraction with super.AI.
- One document or a batch? For one file, copy and paste or Google Docs is enough. For hundreds, use a library or an API.
- Native or scanned? Native PDFs work with most extractors. Scans need OCR.
- Text or fields? For searchable text, any method works. For fields in a spreadsheet or system, plan for field extraction and review.
- How sensitive are the files? Keep confidential documents offline, or use a service whose security terms you have reviewed.
- How much developer time do you have? Libraries are free but need code and upkeep. Cloud APIs need an account and code. No-code tools need neither but do not scale to large batches.
#Frequently asked questions
Yes, but you need OCR, because a scanned PDF stores pictures of pages, not characters. Free options include Adobe's online OCR tool and opening the file with Google Docs in Drive. Developers can use Tesseract or a cloud OCR API. Check numbers and names against the original.
Yes. Copy and paste works on any PDF with a text layer. Google Docs, Adobe's online OCR tool, and Preview on a Mac are free to use, and pdftotext, pypdf, pdfplumber, and Tesseract are free to download. Free online converters work too, but your file goes to a third party.
PDFs store each piece of text with a position on the page, not in reading order, so extractors have to guess. Multi-column layouts, sidebars, and tables are the usual causes. Try another mode, such as pdftotext with and without -layout, or PyMuPDF's sort option.
Plain text extraction flattens tables, so use a tool built for them. In Python, pdfplumber's extract_tables() returns the text of each table as rows of cells, and it works best on native PDFs. Scanned tables need OCR plus table detection. Check merged cells and rows that continue onto the next page.
#What is the difference between a PDF parser and OCR?
A PDF parser reads the text and structure a PDF already contains: characters, fonts, and positions. OCR reads characters from an image of a page. A native PDF needs only a parser, while a scan needs OCR. Some tools do both, such as Google's Enterprise Document OCR, which extracts embedded text from digital PDFs and uses OCR otherwise.
Yes. OCR engines and cloud APIs use machine learning to read printed text, and some also read handwriting. AI models can return specific fields, such as an invoice total, instead of raw text. AI output can still be wrong, so for business data, use a tool that flags uncertain values for a person to check.
#Try super.AI on your own PDFs
If you need text from one PDF, the free methods above will do. If you need the same fields from many PDFs in a spreadsheet or another system, try super.AI Flows. Upload a file, describe what you need in plain language, and super.AI builds the flow and shows each step. You can review each step before you run it. You can upload PDFs, images, and email attachments, and outputs include CSV, JSON, and Google Sheets. For enterprise teams, super.AI offers SOC 2 Type II and contractual commitments. It is free to start, with no credit card and no sales call required. Start for free or book a demo.
#Sources
- Adobe, Convert PDFs to text and XML formats in Acrobat: https://helpx.adobe.com/acrobat/desktop/save-export-documents/convert-to-other-formats/pdf-to-text.html, checked October 15, 2026.
- Adobe, OCR a PDF: https://www.adobe.com/acrobat/online/ocr-pdf.html, checked October 15, 2026.
- Google, Convert PDF and photo files to text: https://support.google.com/drive/answer/176692, checked October 15, 2026.
- Microsoft, Edit a PDF: https://support.microsoft.com/en-us/topic/b2d1d729-6b79-499a-bcdb-233379c2f63a, checked October 15, 2026.
- Apple, Select and copy text in a PDF in Preview on Mac: https://support.apple.com/guide/preview/prvw1020/mac, checked October 15, 2026.
- Apple, Interact with text in a photo using Live Text in Preview on Mac: https://support.apple.com/guide/preview/prvw625a5b2c/mac, checked October 15, 2026.
- Debian manpages (poppler-utils), pdftotext(1): https://manpages.debian.org/trixie/poppler-utils/pdftotext.1.en.html, checked October 15, 2026.
- pypdf, Extract Text from a PDF: https://pypdf.readthedocs.io/en/stable/user/extract-text.html, checked October 15, 2026.
- pdfplumber, README on GitHub: https://github.com/jsvine/pdfplumber, checked October 15, 2026.
- PyMuPDF, Text: https://pymupdf.readthedocs.io/en/latest/recipes-text.html, checked October 15, 2026.
- PyMuPDF, OCR: Optical Character Recognition: https://pymupdf.readthedocs.io/en/latest/recipes-ocr.html, checked October 15, 2026.
- Apache Software Foundation, Apache PDFBox: https://pdfbox.apache.org/, checked October 15, 2026.
- Apache Software Foundation, Apache Tika: https://tika.apache.org/, checked October 15, 2026.
- Tesseract OCR, README on GitHub: https://github.com/tesseract-ocr/tesseract, checked October 15, 2026.
- AWS, Detecting Text (Amazon Textract): https://docs.aws.amazon.com/textract/latest/dg/how-it-works-detecting.html, checked October 15, 2026.
- AWS, Lines and Words of Text (Amazon Textract): https://docs.aws.amazon.com/textract/latest/dg/how-it-works-lines-words.html, checked October 15, 2026.
- AWS, Set Quotas in Amazon Textract: https://docs.aws.amazon.com/textract/latest/dg/limits-document.html, checked October 15, 2026.
- AWS, Amazon Textract pricing: https://aws.amazon.com/textract/pricing/, checked October 15, 2026.
- Google Cloud, Enterprise Document OCR: https://docs.cloud.google.com/document-ai/docs/enterprise-document-ocr, checked October 15, 2026.
- Google Cloud, Document AI pricing: https://cloud.google.com/document-ai/pricing, checked October 15, 2026.
- Microsoft Learn, Document Intelligence read model: https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/concept-read, checked October 15, 2026.
- Microsoft Learn, What is Azure Document Intelligence in Foundry Tools?: https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/overview, checked October 15, 2026.
- Microsoft Azure, Azure Document Intelligence pricing: https://azure.microsoft.com/en-us/pricing/details/ai-document-intelligence/, checked October 15, 2026.
