Our tool analyzes the metadata of a .docx-file, in order to unravel indications of academic misconduct
This interpretation only works if the document had regular saving points (i.e., auto-save was active) while it was written; otherwise, copying the text into a fresh .docx file removes this metadata, making the analysis meaningless.
Got your word file stores in MS Drive or Google Drive? Try our Version History Tool to track the evolution of your documents
To obtain a better analysis, you should use Mentafy to its full extent for your project right from the start. Read More
- Key metadata: Total number of words, words per revision, last modified by, and created by.
- Words per Revision: The average number of words introduced into the document per revision.
- Visual Analysis: A detailed, interactive visualization of the document's metadata. Each row corresponds to the entire document, and hovering over a colored block reveals the associated text passage.
A row changes colors as soon as text came in at a different edit
Each color represents a different point in time, at which words came into the document
A row should always have more than one color.
A human-written text usually has a significant amount of revisions, so you should also always see regular ‘sparkles’ within each row.
If a line with a single color is longer than one row, this line is very likely copied into the documet and NOT written by a human on their keyboard.
Work with Mentafy from the beginning of your writing, because
We will generate a concise report with indicative thresholds, which allows you to judge authenticity within a few minutes and generate a concise report with indicative thresholds, which allows you to judge authenticity within a few minutes.
We provide the authors with writing support from the first steps of finding a topic until the final submission. Most importantly, real-time feedback for the students, encouraging them to work diligently and honestly.
We track document changes reliably (Note: Even if you move text around within a document, we keep the information, how it came in originally) We are not keyloggers, and we respect our users privacy.
FAQ
The tool reads the metadata embedded in a Word Document (.docx) — specifically the revision history that Word stores internally when auto-save is active. From this it reconstructs the chronological order in which text entered the document: how many words were added per revision, when edits occurred, and whether the text grew organically over time or arrived in large, sudden blocks. It does not read or store the content of your document.
Yes to both. Word's internal metadata only captures meaningful revision points when the document is saved regularly — which in practice means auto-save must be switched on while writing. Beyond that, for the full version history to be available, the document also needs to be stored on OneDrive or Google Drive during the writing process, and that cloud version history must be shared as part of the submission. A document written offline or saved only at the end will contain far too few revision points for a meaningful analysis.
Each colour in the visualisation represents a distinct point in time at which words entered the document. A long unbroken block of a single colour means a large amount of text arrived in one revision — characteristic of text that was copied and pasted rather than typed. The longer the single-colour block relative to the overall document, the stronger the signal. Human writing almost never produces this pattern at scale.
It looks like a rich patchwork of colours. Humans write, revise, delete, and rewrite — meaning text enters a document in many small increments across many revisions. A genuinely human-written document will show frequent colour changes within each row, reflecting the natural rhythm of composition. Regular "sparkles" of change are a positive sign of organic authorship.
Not directly on its own — and it is important to be precise about this. The Word Document Analyzer detects how text entered the document: was it typed gradually (human-organic), pasted in bulk (copy-paste), or typed quickly in large chunks (copy-typed)? It does not itself classify text as AI-generated. That verdict comes from Mentafy's AI Detection tool, which analyses the statistical patterns in the text itself, and from the Plagiarism Checker, which identifies whether a source exists. Used together, all three tools provide a much stronger and more evidenced integrity assessment than any single tool alone.
Yes — and we are transparent about this limitation. If text is copied into a fresh Word Document, the original revision history is lost, and the analysis becomes meaningless. This is precisely why the tool works best as part of Mentafy's full workflow: when students write within Mentafy or in a cloud-connected document from the outset, the revision history is preserved and cannot be retroactively erased. The Version History Tool — which uses cloud version history from OneDrive or Google Drive — is significantly harder to circumvent for the same reason.
Both reconstruct how a document evolved, but from different data sources and with different depth. The Word Document Analyzer reads the metadata stored inside the .docx file itself — quick to use, but limited to what Word's internal auto-save captured. The Version History Tool reads the full cloud version history from OneDrive or Google Drive, which is far more granular, harder to manipulate, and provides the basis for Mentafy's full Authorship Report. If you have access to the cloud version history, that is always the stronger analysis.
Strictly Word Documents (.docx). Other formats — including older .doc files, .odt, or PDFs — are not supported, because the revision metadata that the analysis depends on is specific to the .docx format. If a student submits in a different format, ask them to save a copy as .docx in Word before uploading — though note that converting from another format may affect the revision history.
Yes. Mentafy is designed from the ground with GDPR/CCPA principles in mind (data minimization, purpose limitation, user controls). Learn more >