Remove hidden data and metadata

InspectHiddenData() lists the metadata and hidden data in a document and changes nothing. RemoveHiddenData(options) removes the selected categories and returns the same report, with an outcome on every finding. The categories follow what Office's Document Inspector checks, and the same report covers Word, Excel, PowerPoint, Visio, OpenDocument, EPUB, PDF and JPEG, PNG, WebP, GIF and TIFF images.

Clean a file

using OfficeIMO;

OfficeHiddenDataReport report = OfficeDocument.RemoveHiddenData("Contract.docx", "Contract-shared.docx",
    new OfficeHiddenDataOptions {
        Kinds = OfficeHiddenDataKind.Default | OfficeHiddenDataKind.HiddenContent,
        TrackedChanges = OfficeTrackedChangesAction.AcceptAll,
        AnonymousAuthor = "Reviewer"
    });

foreach (OfficeHiddenDataFinding finding in report.Findings) {
    Console.WriteLine($"{finding.Outcome} {finding.Kind} at {finding.Location}: {finding.Description}");
}

OfficeDocument.RemoveHiddenData in the OfficeIMO package picks the format from the file extension, writes the cleaned copy atomically and never changes the input. The output must keep the input's extension. To work in memory, pass the bytes and a file name instead; the result holds the report and the cleaned bytes:

using OfficeIMO;

byte[] original = File.ReadAllBytes("photo.jpg");
OfficeHiddenDataResult result = OfficeDocument.RemoveHiddenData(original, "photo.jpg");
File.WriteAllBytes("photo-shared.jpg", result.ToArray());
Console.WriteLine($"Changed: {result.Report.Changed}, {result.Length} bytes");

To only look, inspect. OfficeDocument.InspectHiddenData takes a path, or bytes and a file name, and changes nothing:

using OfficeIMO;

OfficeHiddenDataReport found = OfficeDocument.InspectHiddenData("Contract.docx");
foreach (OfficeHiddenDataFinding finding in found.Findings) {
    Console.WriteLine($"{finding.Kind}: {finding.Description} {finding.Value}");
}

OfficeDocument.SupportsHiddenData(path) tells you whether a file name has a supported extension. OfficeDocument.Inspect with IncludeHiddenData = true adds the same report to a full inspection.

On a loaded document

A document you already have open has the same two methods. Save it to keep the result.

using OfficeIMO;
using OfficeIMO.Excel;

using var workbook = ExcelDocument.Load("Budget.xlsx");

OfficeHiddenDataReport report = workbook.RemoveHiddenData();
foreach (OfficeHiddenDataFinding finding in report.GetFindings(OfficeHiddenDataKind.ExternalReferences)) {
    Console.WriteLine($"{finding.Outcome}: {finding.Value} {finding.Note}");
}

workbook.Save("Budget-shared.xlsx");

WordDocument, ExcelDocument, PowerPointDocument, VisioDocument, OdfDocument and EpubPublication have InspectHiddenData() and RemoveHiddenData(options). PdfDocument's versions take a CancellationToken, and its RemoveHiddenData returns an OfficeHiddenDataResult with the rewritten bytes, because a PDF is cleaned in one verified rewrite:

using OfficeIMO;
using OfficeIMO.Pdf;

using PdfDocument pdf = PdfDocument.Load("Report.pdf");
OfficeHiddenDataResult result = pdf.RemoveHiddenData(new OfficeHiddenDataOptions {
    Kinds = OfficeHiddenDataKind.Default | OfficeHiddenDataKind.Attachments
});
File.WriteAllBytes("Report-shared.pdf", result.ToArray());

What each format reports

• Word: properties, comments, tracked changes, reviewer names, hidden text, document variables, custom XML, revision save IDs (RSIDs), the attached template and picture metadata. Comments and revisions that stay keep an author, because Word requires one; PersonalIdentities replaces the name with AnonymousAuthor (Author by default). Bibliography sources and custom XML bound to content controls are kept.
• Excel: notes, threaded comments and their authors, hidden sheets, rows, columns and names, external links, pivot cache source paths, custom XML, picture metadata and printer settings. Cached values from external workbooks are removed, but the links themselves are kept, because formulas and refreshes need them. Power Query data is kept and reported with a note. Deleting hidden sheets, rows and columns shifts or breaks formulas that refer to them, exactly as deleting them in Excel does.
• PowerPoint: properties, classic and modern comments, speaker notes, the thumbnail, custom XML and picture metadata. Hidden slides, hidden shapes and shapes placed entirely off the slide count as hidden content.
• PDF: document properties and XMP metadata, comments and markup, attachments, JavaScript and other actions, bookmarks, optional-content layers and Content Credentials. The title stays in the document information and in XMP, as it does in Office files, because readers and screen readers announce it. The PDF/A, PDF/UA and PDF/X identification also stays, so a conforming file keeps declaring its conformance, together with the document IDs PDF/X requires. Inspection leaves out what always stays, so a cleaned PDF doesn't look unfinished; the removal report lists it as retained with the reason. A Factur-X or ZUGFeRD invoice keeps its e-invoice identification while its invoice XML stays attached; select Attachments to remove both. Tagged PDFs, including PDF/UA files, keep their structure tree, language and reading order; a removed comment or attachment is also removed from the tags. Signed PDFs are refused.
• Images: Exif, GPS, XMP and IPTC metadata, Content Credentials and AI source labels. Pixels are not re-encoded.
• Visio, OpenDocument and EPUB: properties, comments, pictures and their metadata, and provenance, as far as each format carries them.

The hidden data support matrix lists exactly what each category covers in each format.

Choose categories

OfficeHiddenDataOptions.Kinds is a flags value. Default selects every category that leaves visible content alone; All adds the rest.

Pictures and provenance

Picture metadata cleanup and provenance overlap: an AI-generated picture often says so in its XMP metadata, and Content Credentials are bound to the exact bytes of the image. While Provenance is not selected, ImageMetadata keeps a picture's digital source type declaration and removes the rest, and leaves pictures that carry Content Credentials untouched. Those findings say so through Reason. Add OfficeHiddenDataKind.Provenance to remove the records too, and use options.Provenance to choose which kinds of record go.

A picture stored in a package that nothing in the document shows is reported with Reason UnusedImage; cleaning removes the whole picture.

Read the report

using OfficeIMO;

OfficeHiddenDataReport report = OfficeDocument.RemoveHiddenData("Contract.docx", "Contract-shared.docx");

foreach (OfficeHiddenDataFinding finding in report.Findings.Where(f => f.Outcome == OfficeHiddenDataOutcome.Retained)) {
    Console.WriteLine($"Kept {finding.Kind} at {finding.Location} ({finding.Reason}): {finding.Note}");
}

HasHiddenData tells you whether anything was found, Kinds which categories were found, Changed whether removal changed the document, and Format which format reported it. Match on Reason rather than the Note text when your code needs to act on why something stayed.

To save the report as JSON, OfficeHiddenDataReportSerializer.Serialize(report, sourceName, outputName) in OfficeIMO.Workflows writes the officeimo.hidden-data.result.v1 schema that the command line, the MCP tools and the browser tool also use.

Signed documents

Removing hidden data changes the file, which breaks a digital signature. Saving a signed Word, Excel or PowerPoint document is blocked by default. Clean the document first and sign it afterwards. To clean a signed package anyway, pass OfficeSignatureMutationPolicy.RemoveInvalidatedSignatures to OfficeDocument.RemoveHiddenData, or set SignatureMutationPolicy on WordSaveOptions, ExcelSaveOptions or PowerPointSaveOptions; the signatures the change would break are removed. Signed PDFs are always refused.

Command line

officeimo hidden-data inspect .\proposal.docx
officeimo hidden-data remove .\proposal.docx
officeimo hidden-data remove .\proposal.docx --output .\shared.docx --force --kinds default,provenance --tracked-changes accept
officeimo hidden-data remove .\poster.pdf --kinds default,provenance --keep-provenance c2pa

The first command reports everything without changing the file. The second writes proposal.cleaned.docx with the default categories. The third also removes origin records and accepts tracked changes, into a named file. The last removes only the AI label and keeps the Content Credentials.

remove never changes the input. It writes <name>.cleaned<ext> unless you pass --output, and refuses to replace an existing file without --force. --kinds takes default, all or category names such as hidden-content, image-metadata, attachments, active-content, bookmarks and provenance. --keep-provenance keeps some kinds of record when provenance is selected: c2pa, external or ai-source. --anonymous-author sets the replacement name, and --remove-invalidated-signatures allows cleaning a signed package. Output is JSON in the officeimo.hidden-data.result.v1 schema, one entry per finding with its outcome; --format text prints a readable list instead.

PowerShell

Get-OfficeHiddenData -Path .\proposal.docx
Remove-OfficeHiddenData -Path .\proposal.docx
$options = New-OfficeHiddenDataOptions -Kind Default, Provenance -TrackedChanges AcceptAll
Get-ChildItem .\Outgoing -File | Remove-OfficeHiddenData -OutputDirectory .\Share -Options $options

Remove-OfficeHiddenData writes <name>.cleaned<ext> unless you pass -OutputPath or -OutputDirectory, replaces an existing output unless you pass -NoClobber, and returns the findings with their outcomes. Clean files before sharing shows real output for Word, Excel, PowerPoint and PDF files.

For AI agents

The OfficeIMO MCP server has the same two operations:

• officeimo_inspect_hidden_data reports what would leave with a file and changes nothing.
• officeimo_remove_hidden_data writes a cleaned copy to outputPath. It takes kinds, trackedChanges (keep, accept or reject), keepProvenance, anonymousAuthor, removeInvalidatedSignatures and overwrite, with the same meanings as the command-line options.

In the browser

Clean a file before sharing runs InspectHiddenData and RemoveHiddenData in your browser. It lists each category it found with the Default categories ticked, and saves a clean copy after checking it again. The file is not uploaded.