The TikTok AI-generated label and how to post without it
TikTok labels videos and slideshows automatically when the file carries Content Credentials. What it reads, which apps write it, and how to post without the tag.
4 min read
On this page7 sections
Short answer. TikTok applies its "AI-generated" label in two ways: you switch on the AI-generated content toggle when posting, or the file carries C2PA Content Credentials and TikTok labels it for you. The automatic label is about the file. Remove the Content Credentials and AI metadata, re-save the video or images, and TikTok has nothing to read. The video remover and the image remover do that, first file free. Realistic AI content still has to be disclosed under TikTok's rules.
Two labels that look the same
TikTok shows one label, but there are two routes to it.
The creator toggle. When you post, the AI-generated content option sits in the post settings. TikTok requires it for realistic AI content: faces, voices, scenes that could be mistaken for real footage. Content you flag this way is labeled, and TikTok says unlabeled realistic AI content can be removed.
The automatic label. Since 2024 TikTok has read C2PA Content Credentials on upload. If the manifest says the content was generated or edited with AI, the label goes on without asking you. TikTok also attaches its own Content Credentials to videos made with its in-app AI effects, so those are labeled too.
Most "why is my video labeled" questions are about the second route, and the answer is in the file.
What TikTok reads
C2PA Content Credentials are a signed block of metadata inside the file. In an MP4 or MOV it sits in a uuid box next to the video tracks; in a JPEG or PNG it sits in a JUMBF segment. The manifest names the tool that made the file and records what it did, including an "AI generated" source type. What is C2PA? explains the format.
Tools that write it into video include Sora, Veo (in the Gemini app and Google Flow), Adobe Firefly video and Premiere's AI features. For images: ChatGPT, Gemini and Nano Banana, Firefly, Photoshop's generative tools, Microsoft Designer and Samsung Galaxy AI.
Chinese generators such as Kling, Hailuo and Doubao write a different mark, an AIGC label required by China's labeling rules. TikTok has not said it reads those, but the sister app Douyin does, and the label still names the tool to anyone who opens the metadata.
Slideshows are checked image by image
Photo mode posts are the fastest-growing AI format on TikTok, and each image in a slideshow is read on its own. One image with Content Credentials labels the post. If you build slideshows from Gemini or ChatGPT images, clean the whole set. The image remover handles 20 images in one batch and gives each its own invisible variation, so the set does not share a fingerprint either.
How to post without the automatic label
- Check the file. Drop it into the free AI label checker to see the Content Credentials, the AI source type and the generator name TikTok would read. It runs on your device.
- Remove the metadata. Upload the video to the video remover (MP4, MOV or WebM up to 150 MB) or the images to the image remover. The C2PA manifest and every AI tag come out, and the file is re-saved with camera metadata from a real iPhone.
- Post the clean file as a new upload, with the AI-generated toggle set the way TikTok's rules require for your content.
The video is also re-encoded with a small zoom, a tiny speed change and a timing offset. Viewers see the same clip; TikTok's duplicate matching sees a new one, which matters if you post the same video to more than one account.
TikTok's rules still apply
TikTok's Community Guidelines require a label on AI content that shows realistic scenes or people. That rule is about what viewers see, not what the file says, and removing metadata does not change it. For stylized content, animation, product shots and illustrations, the label is your call. For realistic content, the honest choice is the toggle.
What does not change
- A visible watermark stays. Sora's moving logo and the Veo mark are drawn into the frames. Metadata removal does not touch pixels you can see.
- SynthID stays. Google's invisible watermark in Veo and Gemini output is built to survive re-encoding. HideSlop makes no claim about it. See the Veo watermark guide.
- A label you applied stays on that post. Post the clean file again with the toggle off.
FAQ
Does TikTok label Sora videos automatically?
Sora writes C2PA Content Credentials into every export, and TikTok reads them, so yes. Remove the credentials and the automatic label does not trigger. Sora's visible watermark is separate and stays in the picture.
Can I remove the AI-generated label after posting?
TikTok offers no way to remove an automatic label from a published video. Delete it, clean the file, and upload it again.
Does TikTok strip metadata from my video?
TikTok transcodes every upload, so the copy viewers get has no metadata. It reads the original first, which is when the label decision happens.
Does the AI-generated label reduce views?
TikTok has not published figures. Many creators report lower reach on labeled posts, and TikTok lets viewers choose to see less AI content, which affects labeled posts directly.
Does this work for Instagram Reels too?
Yes. Meta reads the same Content Credentials plus the IPTC AI field. The cleaned file posts without the automatic label on both. The Instagram specifics are in how to remove the AI label from an Instagram post.