Ghosts in the cargo hold.
AI Metadata Cleaner#
File-format-level metadata removal, independently verified — not a black box#
This page covers exactly one category from the broader AI watermark remover overview: embedded file metadata in documents and images. If you need PDF, JPEG, PNG, or WebP files stripped of author/tool/location fields and want to actually confirm they're gone (not just trust a progress bar), this is the relevant mechanism.
Supported formats and fields#
| Format | Fields removed | How |
|---|---|---|
DocInfo dictionary (/Info: Author, Producer, Creator, etc.), XMP metadata stream |
pikepdf object-graph editing | |
| JPEG | EXIF (including GPS coordinates, camera/tool info), XMP, IPTC, comment segments | Direct segment deletion, no re-encoding |
| PNG | EXIF (eXIf chunk), XMP/text chunks, tIME |
Direct chunk deletion, no re-encoding |
| WebP | EXIF and XMP RIFF chunks | Direct chunk deletion, no re-encoding |
GPS coordinates aren't a separate feature — they live inside the EXIF segment, so removing EXIF removes them as part of the same operation.
Why byte-level matters#
MarkMyAss's image cleaning works at the byte/segment level, not by decoding and re-encoding the image. That means:
EXIF removed
XMP removed
ICC color profile preserved
Pixel dimensions unchanged
Visual content unchanged (byte-identical pixel data)
A tool that decodes and re-encodes your image to "clean" it is doing more than removing metadata — it's also silently altering compression and potentially quality. MarkMyAss doesn't do that: pixel data is never touched, only the metadata segments/chunks around it.
PDF cleaning works the same way conceptually: pikepdf edits the
document's object graph directly, so pages, fonts, images, text, and
links are untouched — only /Info and the XMP metadata stream are
removed. MarkMyAss reopens the cleaned PDF and confirms it's still
structurally valid before handing it back.
Independent verification with ExifTool#
MarkMyAss doesn't ask you to trust its own "removed" claim. If
ExifTool is installed (it is, automatically, in
the hosted deployment), ghostmark verify re-scans the cleaned file with
that separate, independently maintained tool and reports whether it
agrees:
ghostmark clean photo.jpg
ghostmark verify photo.ghostmark.jpg --original photo.jpg
MarkMyAss verification: PASS
ExifTool verification: PASS
Overall: VERIFIED CLEAN
Every property ExifTool reports is categorized (embedded metadata vs. structural/filesystem/computed information), so a preserved ICC profile or the file's byte size is never mistaken for "metadata MarkMyAss failed to remove." Full methodology: /lab/pdf-metadata.
What this page doesn't cover#
This page is about file-format metadata specifically. It says nothing about statistical text watermarks, C2PA provenance manifests, or hidden Unicode — those are separate mechanisms with their own pages: /c2pa-remover, /hidden-unicode-remover, and the broader /ai-watermark-remover overview.