Platforms Cheatsheet: What Each Network Looks At
Quick reference: which metadata blocks each major social platform reads, and the suppression patterns associated with them.
Quick reference
This is the field-tested matrix for which metadata signals each major visual platform reads as of 2026. Suppression patterns are based on public algorithm disclosures, creator-side reach analytics, and platform transparency reports.
| Platform | Reads EXIF Software | Reads XMP CreatorTool | Reads C2PA / JUMBF | Reads filename | Visible "Made with AI" tag |
|---|---|---|---|---|---|
| yes | yes | yes | yes | yes | |
| yes | yes | yes | strict | partial | |
| yes | yes | yes | yes | yes | |
| TikTok | yes | partial | yes | yes | yes |
| partial | partial | yes | partial | yes | |
| X (Twitter) | partial | no | partial | partial | partial |
| Threads | yes | yes | yes | yes | yes |
| YouTube (thumbnails) | partial | partial | yes | partial | yes |
Notes per platform
Instagram & Threads
Same backend. Aggressive about C2PA - even a stripped EXIF file with an intact JUMBF chunk is enough to trigger the "Made with AI" badge and reduce explore-tab eligibility.
The harshest platform for AI signals. They specifically deprioritize filenames matching known AI tools, and their crawler reads C2PA manifests during the pin enrichment step. A pin with both an AI filename pattern and an embedded JUMBF block can be filtered out of the home feed almost entirely.
TikTok
Reads metadata on still uploads (carousel posts and image stories). Does not currently appear to penalize pixel-only images that lack detectable AI signatures.
Less aggressive overall but does respect C2PA. Most reach issues on LinkedIn come from generic engagement patterns, not metadata - though explicit AI tags can dampen reach in the Creator program.
X (Twitter)
Lighter-touch on AI metadata as of 2026. Still recommended to clean files because grok-driven content moderation reads them, and third-party clients sometimes display credentials.
What to clean for each
If you only have time to do one thing per platform:
- Pinterest: randomize the filename. Their pattern matcher is
unusually strict.
- Instagram, Threads, Facebook: strip C2PA. The JUMBF block is the
primary trigger.
- TikTok carousel posts: strip both EXIF and XMP
Softwaretags. - X / LinkedIn: general clean is enough; no platform-specific
hot-spot.
ScrubAI's default settings handle all of the above in one pass.