Document Metadata and Authorship: What Your .docx and PDF Files Actually Record
Every essay file carries a second, invisible document inside it. Document metadata is the set of properties a word processor or PDF tool writes into the file: an author name, a creation date, a last-modified date, the application that made it, sometimes a revision count and minutes of editing time. Students discover it when an instructor mentions it, usually after a detector flag. The honest answer is that metadata alone proves very little, that it corroborates a great deal when it agrees with drafts and version history, and that a standard called content credentials is changing what provenance can mean.
In short: Document metadata records who and when according to the software, and any of it can be edited, stripped or reset by a copy or an export. Alone it is weak evidence of authorship. Beside dated drafts and version history that tell the same story, it becomes a useful cross-check. Content credentials may eventually add tamper-evident provenance to documents.
What document metadata records in a .docx file
Microsoft's own description is the clearest starting point. Document properties, which Microsoft says are "also known as metadata," include standard fields such as title, author name, subject and keywords, which the user can set, and a second group of automatically updated properties the user cannot change directly: file-system details such as the dates a file was created or last changed, plus statistics such as word count. A .docx file is a compressed bundle of XML parts, and these properties live in two small files inside it: core properties (creator, created, modified, last modified by, revision number) and application statistics (pages, words, total editing time).
To see them in Word, open the File menu, choose Info, and read the Properties panel; Show All Properties expands it. Outside Word, right-click and Properties on Windows, or Get Info in the Mac Finder, shows the file-system dates only.
What a PDF records, and how to see it
A PDF carries a similar set: title, author, subject and keywords; the creating application (a field usually labelled Creator, which might say Microsoft Word or Google Docs) and the software that wrote the PDF itself (Producer); and creation and modification dates. Many PDFs also embed a richer metadata block in a format called XMP. In Acrobat or Acrobat Reader, File and then Properties shows the Description tab; Preview on a Mac shows the same under Tools and Show Inspector.
Two things are worth noticing. The Creator field records the application, so a PDF exported from Google Docs says so, which matters if your claim is that you wrote the essay in a Google Doc with version history. And the PDF's creation date is the moment of export, typically minutes before submission; it says nothing about when the writing happened.
Why document metadata is weak evidence on its own
Stevenson's Dr Jekyll and Mr Hyde has a scene that explains the problem better than any manual. Utterson's clerk, Guest, an expert in handwriting, lays Hyde's note beside one from Jekyll and reports that "the two hands are in many points identical: only differently sloped." The observation is correct and the inference is wrong, because Guest assumes a hand belongs to one person. Document metadata is a hand: an artifact of the writing that looks like a signature and is produced by whoever holds the pen. Our guide to Dr Jekyll and Mr Hyde follows the novel's obsession with documents, seals and signatures.
Concretely, the author field is whatever name was typed into the application at setup, often a parent, a previous owner or the word "User." The created date resets when a file is exported or saved as a new file, and it can be set by hand. Microsoft even ships a tool for removing this data: the Document Inspector, under File, Info and Check for Issues, which its documentation says "helps you find and remove hidden data and personal information," warning that removed data cannot always be restored. A field any owner can erase in two clicks cannot carry an authorship claim by itself.
It also cuts the wrong way. A generated essay pasted into a fresh Word file acquires perfect metadata: the student's name, today's date, a plausible edit time. An instructor doing the job well will not treat clean properties as proof, and neither should you. Our guide to how to prove you didn't use AI ranks the evidence that does carry weight.
Why it becomes strong beside drafts and version history
Evidence gains strength from agreement. A single dated draft can be faked; a single metadata field can be faked; a version history can be started fresh. Faking all three so that they agree, session by session, is a different order of effort, and not one a student who generated an essay the night before has made. The value of file metadata is corroboration: where it tells the same story as the rest of the trail, the trail as a whole becomes hard to doubt.
| Field | What it records | How easily it changes | Weight alone | Weight with drafts and history |
|---|---|---|---|---|
| Author or Creator name | The name configured in the application | Editable in seconds; often wrong to begin with | Very low | Low; useful only if consistent across files |
| Created date | When this file object was made | Resets on export or Save As | Low | Moderate if it matches the first draft's date |
| Modified date | Last save | Changes on every save | Low | Moderate if it matches the final session |
| Revision number, editing time | Saves counted and minutes the file was open | Inflates when a file is left open | Low | Moderate; long editing time agreeing with many sessions is telling |
| Creating application (PDF Creator) | Which program produced the file | Fixed at export | Low | Moderate; confirms the workflow you describe |
Read the table as a habit. Write in one document from outline to final, keep it in a cloud folder so version history exists, export the PDF from that document, and the metadata will agree with everything else without effort. Our walkthrough of Google Docs version history as proof of authorship shows what the stronger half of that pairing looks like, and our piece on essay drafts as evidence covers the rest.
Content credentials: provenance that resists tampering
The weakness of ordinary document metadata is that nothing vouches for it. Content credentials are an attempt to fix that. They are defined by an open standard from the Coalition for Content Provenance and Authenticity (C2PA), and they work by attaching a cryptographically signed record, called a manifest, to a file. The manifest lists where the asset came from and what was done to it, and because each part is hashed and signed, any later alteration invalidates the signature. The Content Authenticity Initiative, which promotes the standard, calls the record tamper-evident and persistent across editing, and points to cameras such as the Leica M11-P that write credentials at capture.
Images and video came first because misinformation drove the effort, but documents are on the map. The C2PA specification lists PDF 1.7 and PDF 2.0 among the formats that can embed a manifest, and notes that a plain text file cannot, so a credential for text would have to travel separately. The standard also defines a vocabulary of recorded actions, so a manifest could state that a passage was inserted or that a generative tool was involved. Whether school word processors adopt any of this is open; for now, no common essay workflow produces content credentials.
One caution from the standard itself. C2PA is explicit that credentials do not judge whether provenance information is true, only whether it is well formed, untampered and signed by a trusted party. Hamlet, forging the letter that sends Rosencrantz and Guildenstern to their deaths, stamps it with "my father's signet," the model of the Danish seal, and it passes because the seal is genuine even though the author is not. Our guide to Hamlet reads that scene closely; a valid seal proves the seal, and nothing more.
What a student should and should not rely on
Rely on consistency. Keep the essay in one document with version history, keep the outline and notes, and export the final from the same file, so that every date and name agrees. Know how to open the properties panel before a meeting, so you can show the fields rather than describe them.
Do not rely on metadata to carry an argument alone, and do not edit it after the fact, even to correct a wrong author name. A field that was wrong and stays wrong is an oddity; a field changed after a flag is an accusation waiting to happen. Do not strip metadata from a submitted essay either. And do not assume an instructor reads the fields at all; those who do use them as a cross-check, which is what our guides on authorship and integrity recommend you do as well.
Frequently asked questions
Can a teacher see who wrote a document from its metadata?
They can see the author name the application recorded, which is whatever name was configured on that copy of the software. It is not checked against anyone's identity and the owner can edit it, so it identifies a setup rather than a writer.
Does the created date prove when I started writing?
Only if the file was never exported or saved under a new name since, and only if the computer's clock was right. Treat it as a claim to check against version history and dated drafts rather than as proof.
Should I remove personal information from an essay before submitting it?
Not for graded work. The Document Inspector exists for sharing files with strangers; stripping properties from a school essay removes corroborating evidence and can look evasive.
Will content credentials tell an instructor whether I used AI?
Not today. The C2PA standard supports PDF and can record actions, including generative ones, but ordinary word processors do not yet write credentials, and a credential records what the signing software saw, never a verdict about the writer.
Document metadata is a cross-check, and that is enough
Document metadata records what the software saw: a name, some dates, an application, a count. It cannot see the reading, the outline or the rewrite, and anyone can alter it, which makes it weak alone and useful in company. Keep the trail that matters, let the properties fall into line behind it, and if content credentials ever reach the essay, they will only make the same principle official.
The readings and sample analyses here are companions for your own work; they are not essays to submit.
Sources
- Microsoft Support, View or change the properties for an Office file: document properties "also known as metadata," the standard fields, and the automatically updated properties such as created and changed dates.
- Microsoft Support, Remove hidden data and personal information by inspecting documents: the Document Inspector's purpose and the warning that removed data cannot always be restored.
- C2PA, Content Credentials Technical Specification 2.2: PDF 1.7 and 2.0 among embeddable formats, the external manifest for plain text, and the list of recorded actions.
- Content Authenticity Initiative, How it works: content credentials as tamper-evident, persistent provenance, and camera support such as the Leica M11-P.
- Robert Louis Stevenson, The Strange Case of Dr Jekyll and Mr Hyde, Project Gutenberg and William Shakespeare, Hamlet, Project Gutenberg: sources of the quoted lines.