Tested C2PA metadata on every major social platform. spoiler: its bad

Ran a test uploading C2PA-signed images to every major platform to see who preserves the metadata. Results:

LinkedIn PRESERVES content credentials and actually displays them. only major social platform doing this.
Twitter/X strips everything
Instagram strips everything
Facebook strips everything
Threads strips everything
Bluesky strips everything
Reddit strips everything

so yeah. if you sign your work and post it to social media, the provenance data is just… gone the moment it uploads. everywhere except linkedin.

this is the single biggest barrier to C2PA adoption imo. the standard is technically solid. the camera and software support is growing. but if the platforms where content actually LIVES strip it on upload… whats the point?

the C2PA coalition says theyre “working with platforms” on this. been hearing that for over a year now. when do we start getting frustrated?

The LinkedIn implementation is interesting because it proves it’s technically feasible. It’s not a storage or bandwidth issue — the metadata is tiny compared to the image. The other platforms are actively choosing not to support it.

My cynical take: platforms benefit from the ambiguity. If you can’t prove provenance, you can’t prove copyright violations as easily. And content moderation gets more expensive when you have verifiable creator information attached to everything.

wait does LinkedIn actually DISPLAY the credentials or just preserve them in the file? because theres a difference. preserving = the data survives download. displaying = users can actually see the provenance info on the platform.

if its displaying thats actually huge and I didnt know about it

@zara.phantom both! LinkedIn shows a small “CR” icon on images that have Content Credentials. You can click it to see the provenance chain — who created it, what tool, when, etc. It’s genuinely well implemented.

Which makes the absence on every other platform even more frustrating.

I wonder if the EU AI Act will force this. The labeling requirements for AI content kind of imply that platforms need to support some form of provenance metadata. You cant require labeling if you strip the labels on upload.

Might take regulation to move the needle here because there’s clearly no market incentive for platforms to do this voluntarily (except linkedin apparently).

this is depressing but not suprising. as a photographer ive been signing my images with content credentials for about 6 months now and the only place anyone can actually verify them is my own website and linkedin.

still worth doing though imo. if/when platforms start supporting it, your back catalog is already signed. and for direct client work where you deliver the original file, the credentials are intact.

playing the long game basically.