What PDF metadata can reveal — and what removing it does not remove

Before publishing a PDF, check whether its properties contain a name, internal title or software detail you did not intend to share. Renaming the file does not change those fields.

Open it in the PDF metadata tool. Read the fields found, look for an XMP block, and decide whether you need a cleaned sharing copy. Keep the original if its authorship or archival information matters.

Start with author, title and dates

Author may be a person’s name, an account name or a value inherited from a template. Title, Subject and Keywords can contain an old filename, internal project name or description that no longer fits the document.

Creator usually describes the authoring application. Producer usually names the software that generated the PDF. They can help explain how a file was made, but they are editable fields, not reliable proof of its origin.

CreationDate and ModDate can include a timezone offset. They may be missing, copied from another file or reset during export. The gap between them does not establish how long someone worked on the document, and an offset does not establish where they were.

Check both document properties and XMP

PDFs can use two document-level metadata systems:

• The information dictionary, containing fields such as Title, Author, Creator, Producer and dates.
• XMP, an XML metadata stream that can repeat those fields and add identifiers, rights information or editing history.

A file can have either, both or neither. The values can disagree if an application updated one place but not the other. An empty properties panel does not establish that the rest of the PDF contains no identifying data.

ToolsNow offers Show raw XMP when it finds the catalog’s metadata stream. The preview is limited and may be unreadable for an encoded stream; it is not a complete metadata inventory.

What Remove metadata & download changes

The tool clears the document information dictionary, including custom keys, and empties the XMP stream referenced by the document catalog before removing that reference. It then saves a rewritten copy with -no-metadata.pdf added to the filename stem.

This is a document-level clean. It does not search every object, attachment or embedded image for metadata. If the report says there was nothing to remove, read that as nothing found in the locations the tool handles.

After downloading, open the new file in the inspector again. Check its properties, page count and visible content. For an important publication, use a second PDF inspector as well; repeating the same check can repeat the same blind spot.

Review these parts separately

• Page text and images. A printed name stays on the page. For an image you are sharing separately, use image redaction. For a PDF, use a dedicated content-redaction workflow.
• Comments and annotations. Notes and highlights may include author names and dates. They are not the document’s Author field.
• Attachments. Embedded files can have their own names, content and metadata.
• Forms. Field values and internal field names are not cleared by this operation.
• Embedded objects. Images, fonts, colour profiles and individual PDF objects can carry information outside the two document-level locations.
• Earlier or unreferenced data. Rewriting is not proof that every old object or recoverable value has been removed.

The PDF compressor makes a different kind of copy by rendering pages as images. It loses searchable text and document features; it is not a substitute for reviewing sensitive content or certifying that a file has been sanitised.

When metadata should stay

Metadata supports cataloguing, attribution and rights information. PDF/A archival files require XMP, so removing it can break their conformance. Editing a digitally signed PDF can also affect signature validation.

Accessibility tags are another part of a PDF’s structure, separate from the Author and Title fields. Removing document metadata neither proves that a document is anonymous nor checks that it remains accessible.

How removal can leave hidden data behind

A PDF stores objects connected by references. Deleting a reference to an XMP stream does not necessarily delete the stream’s bytes. That is why this tool empties the referenced stream as well.

For a technical audit, finding an old value in the output is evidence that it survived. Failing to find it with a plain-text search is not proof of removal: PDF strings and streams can be encoded or compressed. A thorough inspection needs a parser and a clear definition of what is being removed.

Sources and further reading

• ISO 32000-2 — PDF specification
• PDF 1.7 specification (ISO 32000-1), §14.3 — document information and metadata streams
• ISO 16684-1 — XMP data model and serialization
• NSA — Hidden Data and Metadata in Adobe PDF Files
• ISO 19005 — PDF/A archival profile
• UK National Cyber Security Centre — Board Toolkit, broader organisational security guidance

Last reviewed 2 October 2026.