The PDF specification defines a fixed set of Document Information Dictionary fields:

**Title** — The document's display name. This is what PDF viewers show in their title bar and what screen readers announce. It should be a human-readable title, not a filename.

**Author** — The person or organisation that created the content. For corporate documents this is often a company name or department.

**Subject** — A short description of the document's topic, similar to an email subject line.

**Keywords** — A comma-separated list of terms for indexing. Search engines and document management systems use these to categorise the file.

**Creator** — The application that originally created the document (e.g. "Microsoft Word 2021" or "Adobe InDesign 2024").

**Producer** — The software that converted the document to PDF (e.g. "Adobe PDF Library 21.0" or "pdf-lib 1.17"). This is often different from Creator when a Word document is exported to PDF.

**CreationDate** — When the PDF was first created, stored as a PDF date string.

**ModDate** — When the PDF was last modified. This updates when you resave the file.

XMP (Extensible Metadata Platform) was introduced by Adobe in 2001 and embedded into PDF 1.4. Instead of a simple dictionary, XMP stores metadata as an XML document inside the PDF's file structure. This XML uses standard namespaces — Dublin Core (dc:), XMP Basic (xmp:), PDF-specific (pdf:) — to describe the same fields and many additional ones.

XMP can carry far richer information than the Document Information Dictionary. A photograph embedded in a PDF might carry EXIF data (camera model, exposure settings, GPS coordinates) as XMP. A scientific paper might carry Dublin Core identifiers like DOI. A scanned document might carry information about the scanner and operator.

Because XMP is extensible, third-party tools can define their own namespaces and embed proprietary metadata without breaking the PDF standard.

PDF metadata is invisible to the reader but fully readable by anyone who opens the file in a tool that exposes it — or by any automated system that processes it. This creates privacy risks that are easy to overlook.

A Word document exported to PDF will carry the author's full name, the company name registered with Microsoft Office, and a revision history. A scanned PDF produced by a photocopier may embed the device's serial number and the user's network login. A PDF produced on a home computer may embed the username from the operating system.

If you are sharing a PDF publicly or with an untrusted recipient, it is good practice to review and clear metadata before sending. Metadata can be stripped using tools that rewrite the Document Information Dictionary and XMP stream with blank values.

Most PDF editors allow you to view and edit the standard Document Information Dictionary fields through a "Document Properties" or "File Info" dialogue. Changing these values does not alter the visible content of the PDF at all — only the embedded descriptive data changes.

Editing XMP metadata requires a tool that understands the XML structure. Adobe Acrobat's File > Properties > Description panel edits both the dictionary and the XMP stream simultaneously. Command-line tools like ExifTool and pdfinfo (poppler) can read and write both.

For SEO and accessibility, the most important field to set correctly is Title — many PDF viewers display it in the browser tab, and screen readers announce it when the document opens.