The properties outside the document body

A DOCX, PPTX or XLSX file can contain properties such as creator, last editor, title, company, application and editing time. Free AI Wiper inspects supported fields in the file's standard document-property parts and separates recognized AI markers from ordinary properties.

This is useful before sharing a draft, deck or spreadsheet when you want to know what the file says about its creation. It does not inspect the prose for AI authorship or change text on a slide, page or sheet.

A careful workflow

  1. Open the cleaner and select a DOCX, PPTX or XLSX file up to 25 MiB.
  2. Read the findings. Choose AI metadata and, if useful, ordinary document properties.
  3. Download the result after the selected findings have been checked again.

For example, removing a Creator property can prevent that property from traveling with a DOCX. It will not remove a name typed in the document itself. Keep the original file if you may need those properties later.

Embedded pictures and thumbnails

The cleaner also inspects supported raster images and SVG drawings in the standard Word, PowerPoint and Excel media folders and document thumbnails. It uses the same metadata rules as the image cleaner, including selective AI and personal metadata removal. Supported formats are JPEG, PNG, WebP, BMP, GIF, TIFF/BigTIFF, still AVIF and supported HEIC. SVG uses the markup cleaner with a 1 MiB per-SVG limit and source-preservation checks. Existing image-layout restrictions still apply. Encoded picture data is preserved.

Verification compares every package entry against the intended edits. Text, formulas, styles and relationships stay unchanged. Custom XML, comments, revision history and opaque embedded files are retained; custom XML may supply visible document content, so deleting it indiscriminately would be unsafe.

Limits

The cleaner works with the supported modern Office formats listed above. Digitally signed Office packages are rejected because changing them would invalidate their signatures. Embedded image inspection is capped at 128 images, 25 MiB per image and 50 MiB together. ZIP inspection is capped at 2,000 entries, 50 MiB per expanded entry and 100 MiB of declared expanded sizes. It has safety limits for unusually large or complex document archives and does not claim to scrub comments, revision history, linked resources or every vendor-specific property. A result with no supported findings is not an AI-authorship verdict.

If the file you want to share is a PDF export, use the separate PDF workflow because its metadata is stored differently.