Explanation
AI labels on your pictures and videos
Which files SemPost marks as made with AI, the Content Credentials it adds, the Made with AI box for uploads, and the label each network gets.
When a picture or a video was made or changed with AI, SemPost marks it in three ways: its card on the Media page says so, the file carries Content Credentials, and a post that carries it goes out with the network's AI label where SemPost can set one. You can't switch this off.
Which files count as made with AI
A file's card on the Media page shows one of three labels next to its size and source:
- AI generated: made by an AI model, such as a scene picture, a motion clip or the clip of a Video Studio shot.
- with AI parts: a picture or video SemPost put together with an AI-made part in it:
- a Studio picture made from a picture with an AI label;
- a Studio video with such a picture, a motion clip, a synthetic voiceover or a track from SemPost's music library that an AI model made;
- a Video Studio video with shots made with AI, a synthetic voiceover, AI ambient sound or such a track;
- a product video from Generate from products with a scene or a motion clip.
- made with AI (marked): a file you uploaded with Made with AI ticked.
The text of a post is never labelled, even when the AI wrote it.
Content Credentials
Content Credentials are a signed record inside the file, under the C2PA standard. SemPost's record says that SemPost made the file and that AI made it, or a part of it. For a file you marked, it says that the person who uploaded it marked it as made with AI.
- A labelled picture gets it when SemPost keeps it, and again when a post carrying it goes out, because the post's shape, text and logo are drawn on the picture then.
- A labelled video SemPost made gets it when SemPost keeps it. A video you uploaded and marked gets it when it's published.
- Each network decides whether it keeps the record when it processes the file.
You can look at a file's record with a Content Credentials reader, such as contentcredentials.org/verify.
Mark your own uploads
SemPost can't tell by itself whether a file you upload was made or changed with AI, so that's up to you Editor. Above the upload area on the Media page, and in the Studio's tray, tick Made with AI before you upload such files. The mark goes with the files you upload next, then the box clears.
- You forgot to tick it. Upload the same file again with the box ticked: SemPost recognises it, keeps it once and marks it.
- You ticked it by mistake. The mark can't be taken off. Move the file to the trash, delete it with Delete now and upload it again without the tick.
What each network gets
When a post carries a labelled file, as one of its pictures, in its collage or as its video, SemPost asks the network for its AI label as the post goes out:
| Network | What happens |
|---|---|
| The post is marked as AI-generated: feed posts, carousels, Reels and stories. | |
| YouTube | The video is marked as altered or synthetic content. |
| The pin is marked as AI-generated. | |
| TikTok | A video published directly is marked as AI-generated content. A post that goes as a draft gets no mark: turn on TikTok's AI label in the TikTok app before you post it. A photo post with a labelled picture always goes as a draft, even when you chose to publish it directly, because TikTok takes no AI mark for photo posts. |
| Facebook, Threads, LinkedIn, Bluesky, Google Business Profile | SemPost can't set a label there. The files still carry their Content Credentials. |
Each network decides how its label looks. Where a network gets no label and you want your audience to know, say it in the post's text.
If something goes wrong
- A TikTok post arrived as a draft though you chose to publish it directly. It carries a picture with an AI label: finish it in the TikTok app and turn on the AI label there.
- A Content Credentials reader says the signer is unknown or not trusted. SemPost signs with its own certificate, which isn't on the readers' lists of trusted signers. The record is still in the file.