PDF Text Extractor - Text & Metadata by URL

PDF Text Extractor extracts the full text of any PDF you give it by URL and returns one clean row per file. You get the text, the page count, word and character counts, the document's own metadata (title, author, producer, creation date) and a sha256 of the file, ready to export as JSON, CSV or Excel or to pull through the API.

Use it when you have a list of PDF links and you need their text in a table: research papers for a RAG pipeline, public filings, standards documents, manuals, reports or forms.

Key features

• Bulk PDF to text. Paste one URL or thousands. Each PDF becomes one row.
• Full text, in reading order. The whole text layer of every page, joined with page breaks.
• Per-page text on request. Turn on includePageTexts to also get a pages array, one entry per page, which is the shape most chunking and embedding code wants.
• Document metadata. Title, author, subject, keywords, creator, producer, creation and modification dates in ISO 8601, and the PDF version.
• File fingerprint. File size in bytes and a sha256 of the exact file that was read, so you can detect when a document changes.
• You pay only for PDFs that yield text. A dead link, a page that is not a PDF, a password-protected file and a scanned PDF with no text layer are skipped, listed with the reason and never charged.

PDF text extractor: how it works

1. Add your PDF links to PDF URLs. Each link must return the PDF file itself.
2. Optionally set Maximum pages per PDF to read only the opening pages, or turn on Include per-page text.
3. Click Start. The Actor downloads each file, reads its text layer and writes one row per PDF to the dataset.
4. Export the dataset as JSON, CSV, Excel or HTML, or read it through the Apify API.

Anything that could not be extracted is written to the SKIPPED record in the run's key-value store, with the URL and the reason. Skipped files are not dataset rows and are not charged.

Input

Field	Type	Default	What it does
pdfUrls	array of strings	required	Direct links to the PDF files to extract.
maxPagesPerPdf	integer	0	Read only the first N pages of each PDF. 0 reads every page. Does not change the price.
includePageTexts	boolean	false	Adds a pages array with each page's text. Does not change the price.
maxFileSizeMb	integer	50	Skips any PDF larger than this. Maximum 200. A skipped file is not charged.

Input example

{
    "pdfUrls": [
        "https://bitcoin.org/bitcoin.pdf",
        "https://www.rfc-editor.org/rfc/rfc9110.pdf"
    ],
    "maxPagesPerPdf": 0,
    "includePageTexts": false,
    "maxFileSizeMb": 50
}

Output

Each row describes one PDF.

Field	Meaning
url, finalUrl	The link you gave and the link the file was served from after redirects.
fileName	The last path segment of the final URL.
fileSizeBytes, sha256	Size and sha256 of the exact file read.
pageCount	Pages in the document.
pagesExtracted	Pages actually read. Lower than pageCount only when you set maxPagesPerPdf.
charCount, wordCount	Size of the extracted text.
text	The extracted text. Pages are separated by a blank line.
textTruncated	true only if a row had to be shortened to fit the dataset's item size limit.
title, author, subject, keywords, creator, producer	The document's own metadata. null when the PDF does not set the field.
creationDate, modificationDate	ISO 8601, or null.
pdfVersion	The PDF format version the file declares.
pages	Only with includePageTexts: [{ "page": 1, "charCount": 2412, "text": "..." }].
extractedAt	When the row was produced.

Sample output data

This is a real row from run gVImysURQHM2DrMS9 of this Actor on 2026-10-03, with the input shown above plus four more URLs. One thing is shortened here: text shows its first two lines of the 21,235 characters the run returned. Every other value is as delivered.

{
    "url": "https://bitcoin.org/bitcoin.pdf",
    "finalUrl": "https://bitcoin.org/bitcoin.pdf",
    "fileName": "bitcoin.pdf",
    "fileSizeBytes": 184292,
    "sha256": "b1674191a88ec5cdd733e4240a81803105dc412d6c6708d53ab94fc248f4f553",
    "pageCount": 9,
    "pagesExtracted": 9,
    "charCount": 21235,
    "wordCount": 3558,
    "textTruncated": false,
    "title": null,
    "author": null,
    "subject": null,
    "keywords": null,
    "creator": "Writer",
    "producer": "OpenOffice.org 2.4",
    "creationDate": "2009-03-24T17:33:15.000Z",
    "modificationDate": null,
    "pdfVersion": "1.4",
    "text": "Bitcoin: A Peer-to-Peer Electronic Cash System\nSatoshi Nakamoto",
    "extractedAt": "2026-10-03T17:58:54.397Z"
}

The same run read a 194-page, 71,309-word standards document and a 6-page tax form into one row each.

Pricing

This Actor uses pay-per-event pricing with one event.

Event	Price	When it is charged
PDF extracted	$0.0055 per PDF ($5.50 per 1,000)	Once for each PDF whose text was extracted and written to the dataset, whatever its page count.

There is no start fee and no per-page fee. A 1-page PDF and a 500-page PDF cost the same.

What is never charged: a download that fails, a URL that does not return a PDF, a password-protected PDF, a file over your size limit, and a scanned PDF with no text layer.

Worked example: 200 PDF links, of which 190 yield text and 10 are dead links, cost 190 × $0.0055 = $1.045.

Each run extracts the PDFs you list in that run. If you run the same link again it is extracted and charged again, so keep a list of what you have already processed and send only new links. The sha256 field lets you check whether a document changed before you re-extract it.

You can set a maximum charge per run in the run options. When the limit is reached the Actor stops, and every row you received is a row you paid for.

FAQ

Does it read scanned PDFs? No. This Actor reads the text layer of a PDF and does not run OCR. A scanned document with no text layer is skipped and not charged. If a PDF mixes scanned and text pages, you get the text pages.

Does it keep tables and layout? It returns plain text in reading order. Table cells arrive as text on lines, without cell structure. Labels inside a diagram can arrive in a different order than they appear on the page.

Can it read a PDF behind a login? No. The link must be reachable without cookies or a password. Some sites refuse automated downloads; those links are skipped with the HTTP status as the reason and are not charged.

How large can a PDF be? Up to the maxFileSizeMb you set, 200 MB at most. Very large documents need more memory; raise the run's memory if a big file fails.

Is it legal to extract text from PDFs? You are responsible for having the right to download and process the documents you submit. This Actor only reads the links you give it. Do not use it on documents you are not allowed to copy, and take care with documents that contain personal data.

How do I use the text with an LLM or vector database? Turn on includePageTexts and send each pages[].text entry to your embedding step, keeping url and page as the citation.

Limits and troubleshooting

• No OCR. Text must exist in the PDF's text layer.
• One row is capped at about 8 MB. A longer text is shortened and textTruncated is set to true.
• A link to a web page that displays a PDF is not a PDF link. Use the direct file URL.
• Three download attempts are made for timeouts and server errors before a link is skipped.

Related Actors

Other Actors from the same publisher that return one typed row per record:

• Open Library & Gutenberg Book Lookup for book metadata by ISBN.
• NPPES NPI Registry Lookup for US healthcare provider records.
• Shopify Store Product Scraper for product catalogues.
• YouTube Channel Upload Monitor for new uploads by channel.

Your feedback

Found a PDF that should have worked? Open an issue on the Actor's Issues tab with the link, and it will be looked at.

Legal

This Actor is unofficial. It is not affiliated with, endorsed by or sponsored by Adobe, the PDF Association, or any site named in the examples above (bitcoin.org, rfc-editor.org). PDF is an open ISO standard (ISO 32000). The example documents are public and are used only to show the output shape.