PDF Duplicate Page Finder

Runs in browser New PDF Analysis & Repair

Drop a PDF and every page is compared with every other: by its text, so a page that differs only by a page number or a date still matches, and by a small picture of the page, for scans without text. Each group lists its page numbers and whether the pages are identical, have the same text, near-identical text or only look the same. Nothing is removed; use Delete PDF Pages afterwards if you want the copies gone.

Privacy: Processed locally in your browser. Your file is not uploaded.

What is inside a PDF file, and what does repairing one do?

A PDF is a header naming its version, a body of numbered objects (pages, fonts, images, streams), a cross-reference table that records where each object starts, and a trailer pointing at the catalogue and the metadata. Metadata lives in the Info dictionary (title, author, dates, producer) and, since PDF 1.4, in an XMP stream; fonts are embedded whole, subsetted or only referenced by name. A viewer that cannot find an object walks the file to rebuild the cross-reference table — repair does the same and writes the rebuilt table, which fixes most "damaged file" errors without touching the pages.

How to use PDF Duplicate Page Finder

  1. Drop a PDF (unlock it if it asks for a password).
  2. Choose how similar the text must be.
  3. Click "Find duplicates".
  4. Read the groups; remove the copies with Delete PDF Pages if you want.

PDF Duplicate Page Finder features

  • Exact duplicates: same text and same appearance
  • Same text with a different look
  • Near-identical text at 95, 90 or 80% similarity
  • Scanned pages matched by appearance
  • Blank pages listed apart
  • Page numbers ready for Delete PDF Pages
  • The file is never changed

PDF Duplicate Page Finder example

A report assembled twice

Input:

report.pdf, 24 pages, threshold 90%

Output:

3 groups: pages 4 and 17 identical; pages 6 and 19 near-identical text (96%); pages 22 and 23 look the same (no text layer)

Frequently asked questions about PDF Duplicate Page Finder

How does it decide that two pages are duplicates?

It reads each page's text layer and a small greyscale picture of the page. The same text and the same picture make the pages identical; the same text with a different look (another layout or a stamp) is "same text"; text that is at least 90% the same — the threshold is adjustable — is "near-identical"; and pages without text are matched by their picture alone.

Does it work on scanned PDFs?

Partly. A scan has no text layer, so its pages are compared by appearance only: the same page scanned twice usually matches, but a second scan that is shifted, rotated or lit differently may not. Run OCR PDF first to add a text layer, then check the searchable copy to compare the text.

Does it remove the duplicate pages?

No. It only reports them, and your file is not changed. It lists the page numbers that repeat an earlier page, which you can type into Delete PDF Pages.

Why are blank pages listed separately?

Blank pages all look alike, so counting them as duplicates would fill the report with noise. They are listed apart, and you can still remove them if you want.

Is there a page limit?

Up to 1,000 pages in one file, because every page is rendered once and near-identical text is compared page by page. Split a longer document with Split PDF first.

Is my PDF uploaded to a server?

No. The file is opened and processed by your browser with WebAssembly and JavaScript; nothing is sent to Mutqan Soft or anyone else. Close the tab and the file is gone. Downloading the result needs a free account, and right now this tool also asks you to sign in before it runs — no subscription, no payment.

Can I use it on a password-protected PDF?

Yes, if you know the password: the file list asks for it once and every operation then works on the decrypted pages. The output is saved without a password unless you protect it again.

Technical notes

Each page's text layer is read with pdf.js and normalised (case, spacing, punctuation; any script). Pages with the same normalised text are grouped, and split into identical and same-text groups by an exact hash of a 32 × 32 greyscale thumbnail rendered from the page. Remaining pages are compared by the Jaccard similarity of their word 3-shingles. Pages with one or two words match only the same text exactly. Pages without any text are compared by the thumbnail hash and a 64-bit difference hash (at most 4 bits apart), and near-uniform ones are reported as blank. Everything runs in the browser, and the document is only read.

References