XLSX forensics

Excel spreadsheets — formula coverage, cell-style anomalies, and worksheet structure.

Excel's OOXML container .xlsx (plus .xlsm for macros) is structurally similar to DOCX and PPTX but the body content is cells and formulas, not prose. The authorship-signal mix shifts accordingly.

Signals available on XLSX

Metadata — same set as DOCX: Creator / Last modified by / Created / Modified / TotalTime / Application / AppVersion. Note: TotalTime is less reliable on XLSX because spreadsheet edit time often happens outside Excel (formulas pasted from elsewhere).
Formula coverage. Authored spreadsheets contain formulas; pasted/exported tables tend to be flat values. The formula-to-value ratio is a strong signal.
Cell-style inventory. Real Excel work uses a handful of named styles; CSV-imported or AI-exported tables tend to be styled uniformly or use the default.
Worksheet structure. The number, naming, and visibility of sheets (hidden sheets are particularly interesting).
Embedded objects. Charts, pivot tables, embedded VBA macros, embedded Word/PowerPoint snippets.
Structural. ZIP-entry inventory; export-source fingerprint (Excel vs. Google Sheets export vs. LibreOffice Calc vs. Pandas / openpyxl programmatic export).

Common false-positive paths

CSV converted to XLSX is structurally identical to a flat AI-exported table. If the assignment allows CSV import, this isn't suspicious.
Google Sheets export trips export-source detection the same way Docs export does.
Programmatic export (Pandas, openpyxl) is common in data-science classes. Recognize the fingerprint before raising it.

What's distinctive about XLSX review

Spreadsheet integrity hinges on the formula structure, not the prose content. The linguistic detectors are far less useful on spreadsheet content. Focus on:

Formula coverage — is this a real model or a flat table?
Sheet-to-sheet references — do the inter-sheet formulas make structural sense?
Hidden sheets — are there sheets the student didn't intend the reviewer to see?
Named ranges and defined names — authored work tends to use named ranges; pasted tables don't.

What to expect

A typical authored XLSX scan surfaces:

Full file metadata.
Sheet-by-sheet inventory.
Formula-to-value ratio per sheet.
Cell-style distribution.
A structural fingerprint of the source tool.

For high-stakes review, request a screen-recording of the student opening the workbook and walking through their work — that's far more informative than static forensics for spreadsheet assignments.