Home Blog PDF Metadata Fields Explained

What Does PDF Metadata Contain? Every Field Explained

Author, Creator, Producer, XMP stream — every field explained with real examples, risk levels, and where each one comes from.

Updated July 2026 · 8 min read
PDF metadata fields explained — Author, Creator, Producer, XMP stream
Direct answer

PDF metadata contains two separate stores: the Info dictionary (Author, Title, Subject, Keywords, Creator, Producer, CreationDate, ModDate) and the XMP stream (an XML block duplicating all of those plus additional fields like revision count and rights data). Most tools only show you the Info dictionary. The XMP stream carries the same personal data and is readable by ExifTool and professional document analysis software.

Open any PDF. Go to File → Properties. You'll see a handful of fields — Author, Creator, maybe some dates. Clean enough. Now run exiftool document.pdf on that same file. Suddenly there are 30 fields, including ones Acrobat never showed you: XMP:CreateDate, dc:creator, xmp:MetadataDate, revision count, rights management strings.

The PDF format has carried metadata since its first commercial version in 1993. It's grown considerably since then. Understanding what's actually in there — and where each field comes from — is the difference between a superficial clean and a complete one.

This is a complete field-by-field reference. Every field, what it contains, where it comes from, and what the privacy risk is.

The two metadata stores — why this matters

Before the fields: the architecture. PDF metadata lives in two places simultaneously, and almost every guide gets this wrong by only discussing one of them.

The Info dictionary is a key-value block embedded in the PDF's cross-reference structure. It's been there since PDF 1.0. Acrobat Reader's Properties panel reads this. It's what you see when you go to File → Properties → Description.

The XMP stream (Extensible Metadata Platform) is an XML document embedded separately in the PDF file body. Adobe introduced it in 2001 and it's now part of the ISO 32000 PDF standard. It contains the same fields as the Info dictionary, plus additional ones — and it's the one that survives when people think they've cleaned a file.

The practical consequence: if you clear the Info dictionary fields in Acrobat's Properties panel and then run ExifTool, your name usually still appears in the XMP stream as dc:creator. Complete removal means stripping both. Our guide on how to remove PDF metadata covers every method that handles both layers.

⚠ The partial clean problem

The majority of online "metadata removers" only clear the Info dictionary. They don't touch the XMP stream. After using them, Acrobat Properties shows blank fields — but ExifTool still finds your name, your software, and your timestamps in the XML layer. This is why verification matters: always check both stores, not just the one Acrobat shows you.

Info dictionary fields — the complete reference

These are the standard fields in the PDF Info dictionary, as defined in the ISO 32000-1 specification. Click any field to expand its full details.

Author The person who created the document
High risk
StoreInfo dict + XMP dc:creator
SourceOS user account / Office profile
Personal dataYes — GDPR Art. 4(1)
Author : Jane Smith
XMP:Creator : Jane Smith

The Author field is automatically populated from your Windows username, Apple ID, Microsoft 365 account name, or Google account display name — depending on the application. You never typed this. Your software inserted it every time you created or saved the document. It is personal data under GDPR Article 4(1) and under UK GDPR. Sending it to external parties without a lawful basis may violate the data minimisation principle. See our guide on how to remove the Author field from a PDF for all removal methods.

Title The document's title
Medium risk
StoreInfo dict + XMP dc:title
SourceDocument properties / filename
Personal dataSometimes
Title : CONFIDENTIAL — Q3 Settlement Draft v4
XMP:Title : CONFIDENTIAL — Q3 Settlement Draft v4

Often set from the document's Title property or pulled from the filename. The risk level depends entirely on what the title says. A title like "CONFIDENTIAL — Client A vs Client B Settlement" is highly sensitive. A generic "Proposal" is not. In Google Docs exports, the Title field is set to whatever the document is named in Google Drive — so internal project codes and client names in Drive filenames end up embedded in the exported PDF.

Subject A description or subject classification
Medium risk
StoreInfo dict + XMP dc:description
SourceManually set or template-inherited
Personal dataRarely
Subject : HR — Performance Review — Confidential

Usually blank unless set deliberately. The risk emerges when templates are reused — a Subject field set in a template for a specific client or project carries forward into every document created from that template. The most common surprise: LibreOffice Writer sometimes auto-populates Subject from document properties inherited from the template file.

Keywords Internal classification tags
High risk
StoreInfo dict + XMP dc:subject
SourceDocument properties / DMS tags
Personal dataSometimes
Keywords : PRIVILEGED, attorney-client, Project-Falcon, draft-v3

This is the field that causes the most damage in legal and professional contexts. Document management systems (iManage, NetDocuments, SharePoint) often write internal classification codes, matter numbers, client codes, and privilege flags into the Keywords field automatically. Those tags travel with the document into every export and every email. A recipient seeing "attorney-client" or "Project-Falcon" in Keywords has just received your internal classification system — which may reveal the nature and status of a matter you haven't disclosed.

Creator The application that created the source document
Medium risk
StoreInfo dict + XMP xmp:CreatorTool
SourceApplication that created the source
Personal dataNo
Creator : Microsoft Word
XMP:CreatorTool : Microsoft Word

Identifies the application that created the original document — Word, InDesign, Google Docs, LibreOffice. Not personal data, but reveals your software stack. This matters when you're charging premium design rates and the Creator says "Microsoft Word 2019", or when you're claiming the document was created in a specific system but Creator contradicts that. It also matters forensically: Creator vs Producer mismatches are a primary indicator of document tampering. See our guide on how to check if a PDF has been edited for the full forensic picture.

Producer The application that generated the PDF
Medium risk
StoreInfo dict
SourcePDF generation library
Personal dataNo
Producer : Microsoft: Print To PDF
Producer : Adobe Acrobat Pro 2024.003
Producer : Foxit PhantomPDF 10.1

Where Creator identifies what made the original document, Producer identifies what converted it to PDF. If you write in Word and export directly, Producer is usually Microsoft's PDF library. If you later open and re-save in Acrobat, Producer changes to Acrobat while Creator stays as Word. This field is central to tampering detection: a document claiming to be an original export from Word, with a Producer showing Foxit PhantomPDF, has been processed by a third tool after initial creation.

CreationDate When the PDF was first created
Low risk
StoreInfo dict + XMP xmp:CreateDate
FormatD:YYYYMMDDHHmmSSOHH'mm'
Personal dataNo
Create Date : 2026:03:14 09:22:11+00:00

Timestamp of when the PDF was first generated. Low personal data risk, but high forensic value. If a document's stated date (printed on page 1) doesn't match its CreationDate, something is off. If CreationDate is much earlier than expected, the document may be a backdated copy. The timestamp also includes timezone offset — which can reveal the geographic location of the machine that created the file.

ModDate When the PDF was last modified
Medium risk
StoreInfo dict + XMP xmp:ModifyDate
SourceSet on each save
Personal dataNo
Modify Date : 2026:07:24 23:47:11+05:45

Updated every time the file is saved or processed. A ModDate after the document's stated effective date proves editing occurred after that date. A ModDate showing 11:47 PM the night before your proposal deadline tells a negotiating counterpart you were under time pressure. For freelancers, a ModDate revealing that a "bespoke" proposal was adapted from a template at 11 PM is not the professional signal you want to send. The timezone offset in this field can also reveal your physical location.

Microsoft Office-only fields

Word, Excel, and PowerPoint add fields beyond the ISO standard when exporting to PDF. These come from your Microsoft 365 or Active Directory profile — data you set up years ago and forgot about.

Field Source Personal data? Risk
Manager Active Directory / Microsoft 365 org chart — your manager's name Yes High
Company Your Office account's organisation field Org name only Medium
Category Document category from Word properties panel Rarely Low
ContentStatus Status from SharePoint or Word — "Draft", "Final", "In Review" No Medium
RevisionNumber Number of times the source Word document was saved No Low

The Manager field is the one that consistently surprises people. If you've ever used Office on a corporate machine — or signed into a corporate Microsoft 365 account — your manager's full name from the Active Directory org chart is sitting in that field. It has been stamping itself onto every document you've exported since the day you logged in. On a personal machine, it usually reflects whoever was listed in Word's user settings when the machine was set up.

See what's in your PDF right now

The Privacy Risk Score shows every field — Info dict and XMP stream — in one view. Free, browser-only.

Open PDF Metadata Remover →

XMP stream fields — the hidden layer

The XMP stream duplicates most Info dictionary fields and adds its own. Here are the XMP-specific fields that don't appear in the standard Info dictionary:

XMP Field Namespace What it contains Risk
xmp:MetadataDate xmp: When the metadata itself was last written — often seconds after ModDate, reveals if metadata was re-processed Medium
xmpMM:DocumentID xmpMM: A unique identifier for the document. Persists across saves and can link different versions of the same document Low
xmpMM:InstanceID xmpMM: Changes on every save — a unique fingerprint for this specific version of the file Low
xmpMM:History xmpMM: Edit history including application used and timestamps for each save (Adobe Creative Suite documents especially) Medium
dc:rights dc: Copyright or rights statement — sometimes contains organisation name Low
pdf:Producer pdf: XMP-specific duplicate of the Producer field Medium

The xmpMM:DocumentID field is particularly interesting from a forensic standpoint. It's a UUID that stays with a document across saves — meaning two files with the same DocumentID are provably derived from the same original. If you share what you claim is an independent document but it has the same DocumentID as a file already in evidence, that's a problem. Conversely, it can be used to prove provenance when that's in your favour.

How to read all metadata fields — verification workflow

Acrobat Reader shows you the Info dictionary. ExifTool shows you everything. Here's the command sequence:

# See all metadata in both stores
$ exiftool document.pdf

# See only the high-risk personal data fields
$ exiftool -Author -Manager -Keywords -Subject document.pdf

# See XMP fields specifically (namespace prefix)
$ exiftool -XMP:all document.pdf

# Compare Info dict vs XMP for same field
$ exiftool -Author -XMP:Creator -ModifyDate -XMP:ModifyDate document.pdf
# Discrepancies between these pairs = partial clean or tampering

GDPR field map — which fields require action

Under GDPR Article 4(1), personal data is any information relating to an identified or identifiable natural person. Under Article 5(1)(c), it must be minimised — limited to what is necessary for the purpose.

For external document sharing, here's which fields require action:

Field Personal data? Action required before sharing
Author Yes — full name Strip or replace with organisation name
Manager Yes — full name Strip — never serves external purpose
Keywords Depends on content Review and strip if containing names or privileged classifications
Title / Subject Depends on content Review — strip if containing client names or sensitive matter codes
Creator No Optional — strip if software disclosure is a concern
Producer No Optional — useful to strip to reduce forensic surface area
ModDate No Optional — strip to prevent editing timeline disclosure
XMP:all Duplicates above Must strip XMP stream — not just Info dictionary

For UK organisations, see our detailed breakdown in UK GDPR vs EU GDPR on document metadata. The obligations are substantively the same but the enforcement authority and some procedural requirements differ post-Brexit.

Field verification checklist

PDF Metadata Field Verification 0 / 6 verified
Check Author in Acrobat Reader — File → Properties → Description
Confirms Info dictionary Author is cleared. If blank here, check XMP separately.
Run exiftool -Author -XMP:Creator document.pdf
Both lines must return empty. If XMP:Creator still shows a name, the XMP stream was not fully cleared.
Check Manager field — exiftool -Manager document.pdf
Often overlooked. Corporate Active Directory users almost always have this populated.
Check Keywords for internal classification tags
DMS systems write matter numbers, client codes, and privilege flags here automatically.
Check Creator vs Producer for tampering indicators
Different applications in Creator vs Producer means the file was processed by a second tool after creation.
Run exiftool -XMP:all document.pdf for full XMP audit
Surfaces all XMP namespace fields including MetadataDate, DocumentID, and edit history not visible in Acrobat.

Scan and strip your PDF in one step

Privacy Risk Score shows every field. One click removes them all. Browser-only — nothing uploaded.

Open PDF Metadata Remover →

Frequently asked

PDF metadata contains: Author (the person who created it), Title, Subject, Keywords, Creator (software used to create the source document), Producer (software that generated the PDF), CreationDate, and ModDate. Microsoft Office also adds Manager and Company from your corporate profile. Modern PDFs store all of this twice — in the Info dictionary and in the XMP stream.
Creator identifies the application that created the original source document — for example "Microsoft Word". Producer identifies the application that generated the PDF output. If you wrote in Word and exported directly, both may say Microsoft. If you later processed the PDF with Acrobat, Producer changes to show Acrobat while Creator stays as Word. A mismatch between Creator and Producer is one of the primary indicators of PDF tampering.
XMP (Extensible Metadata Platform) is an XML-based metadata format embedded as a separate block inside the PDF file body. It was introduced by Adobe in 2001 and is now part of the ISO 32000 standard. PDFs store metadata in two separate locations: the Info dictionary (visible in Acrobat Properties) and the XMP stream (visible via ExifTool). Many metadata removers only clear the Info dictionary, leaving the XMP stream — and your name — fully intact.
The Author field contains the full name of the person who created the document, automatically pulled from the OS or Office account. In Word it comes from File > Options > General > User name. In Google Docs it comes from your Google account display name. You never explicitly typed it — the application inserted it automatically on every save and export. It is personal data under GDPR Article 4(1).
Yes. In Adobe Acrobat Reader (free): File > Properties > Description shows the main Info dictionary fields. On macOS: right-click > Get Info > More Info. On Windows: right-click > Properties > Details tab. For the full picture including XMP stream fields, ExifTool (free command line) or FileIntel's PDF Metadata Remover show everything in one scored view.
Fields containing a natural person's name — Author, Manager — are personal data under GDPR Article 4(1). Sharing documents with these fields populated to external parties constitutes a personal data transfer requiring a lawful basis. Under the data minimisation principle (Article 5(1)(c)), metadata that serves no purpose for the recipient should be stripped before sharing.