Remove PDF Metadata
Read every value the file carries, including custom Info keys and the XMP packet, then clear the groups you choose. The table shows what each group removed, and the copy is written in this browser.
Runs locally in your browser ·
Read every value the file carries, including custom Info keys and the XMP packet, then clear the groups you choose. The table shows what each group removed, and the copy is written in this browser.
Runs locally in your browserDrop a PDF and the page reads the values the file carries: Title, Author, Subject, Keywords, Creator, Producer, the creation and modification dates, and whether an XMP packet or a PieceInfo entry is attached. Any other key in the document information dictionary is listed under its PDF name - a file written as an incremental update showed a Custom row beside the nine standard ones.
Tick the groups to clear and press Remove selected metadata: the table turns into a before/after comparison, the counter reports the values that disappeared, and the cleaned copy downloads as name-metadata-clean.pdf. pdf-lib does the work inside this tab, so the file never leaves the browser.
Document information clears Title, Author, Subject, Keywords and every other key the Info dictionary holds, including a custom key such as /Custom. Software and dates clears Creator, Producer, CreationDate, ModDate and Trapped. Embedded XMP metadata clears the XMP packet and its PieceInfo entry. The two field lists do not overlap, so with only Software and dates ticked the title and the author stay in the copy.
The XMP packet is removed as an object, not merely unlinked: the copy was scanned byte by byte and the marker string the packet carried was no longer in it. Removing the catalog reference alone used to leave the packet readable in the file.
Page count, page content, annotations and form fields travel unchanged: a 40-page file came back with 40 pages, and a form document came back with its AcroForm, its three annotations, the filled-in text and the ticked box.
The trailer keeps its document ID pair, so a file that arrives with an ID leaves with the same one. Clearing metadata does not hide anything that stays visible on a page or is stored elsewhere in the file.
A file that is not a PDF, an empty file and a file cut short all end with the same sentence and no download: the PDF engine reports a truncated file as a document with zero pages, and the page refuses that result instead of writing a copy with an incomplete page tree. An encrypted PDF is reported separately, because the engine cannot decrypt it and the page never asks for a password.
Files edited through incremental updates are read at their newest revision, and older revisions do not come back in the copy: a file whose Info dictionary existed only in the first revision lost its title and author like any other. Strings stored as UTF-16BE (a Chinese title, a Cyrillic author) and both PDF date layouts (D:2024, D:20230101120000-05'30') are read, and dates are shown as ISO 8601 timestamps in UTC.