AI content disclosure labels: what each platform actually requires
AI captions and auto-reframing are not deepfakes, and no platform asks you to label them. Here is where each one draws the line, in its own published words.
9 minute read
The one distinction that decides everything
Every rule below turns on a single question: could a viewer mistake what they are seeing for something that really happened? AI that helps you make a video is not what these policies are aimed at. YouTube's help page lists caption creation, colour and lighting adjustment, background blur, beauty filters, voice or audio repair, and cloning your own voice as minor edits that need no disclosure, on the grounds that they are primarily aesthetic and do not change what actually happened. What triggers a label is synthetic media depicting a real person, place or event in a way a viewer could take as real. Get that line right and most of the compliance work disappears, because almost nothing in a normal clipping workflow crosses it.
YouTube: what the altered content toggle is actually for
YouTube requires disclosure when realistic content is made with altered or synthetic media. Its help page names three triggers: making a real person appear to say or do something they did not, altering footage of a real event or place, and generating a realistic scene that did not occur. You declare it during upload in YouTube Studio, in the Details section under Altered content. The exemptions are equally explicit. Unrealistic or clearly animated content is out of scope, as are the minor edits listed above. YouTube's original 2024 announcement also put productivity uses on the safe side, naming generated scripts, content ideas and automatic captions. Shorts made with YouTube's own generative AI effects disclose themselves, so you do not need to tick anything.
What changed in May 2026: labels moved, and YouTube applies some itself
On 27 May 2026 YouTube announced it had made these labels harder to miss and had started applying some without the creator. For long-form video, the label on photorealistic and meaningfully altered content now sits directly below the player, above the description; on Shorts it appears as an overlay on the video itself. Unrealistic, animated or lightly altered content still gets a description-level disclosure only. YouTube also said its own signals can detect significant photorealistic AI use and apply a label when a creator has not disclosed, and that labels are permanent on content made with its tools such as Veo and Dream Screen or carrying C2PA metadata showing it was fully AI-generated. Incorrect detections can be challenged in YouTube Studio, and YouTube says disclosure does not affect monetisation eligibility or recommendations.
Meta: the AI info label across Instagram, Facebook and Threads
Meta's published approach, first posted in April 2024 and updated since, says it applies an AI info label when it detects industry-standard signals — specifically the AI-generated fields in the C2PA and IPTC technical standards — or when the person posting self-discloses. Meta requires you to use its disclosure tool when you post organic content containing photorealistic video or realistic-sounding audio that was digitally created or altered. The label started life as Made with AI and was renamed after ordinary retouching kept triggering it, which tells you exactly where Meta ended up drawing the line. Where Meta judges that altered media carries a particularly high risk of materially deceiving the public on a matter of importance, it can apply a more prominent label instead of the standard one.
TikTok: the AIGC toggle, and the label you cannot take off
TikTok asks you to label AI-generated or significantly edited media that shows realistic-appearing scenes or people, meaning content that would lead someone to believe the person is real or the event happened. Small edits such as colour correction and cropping are out of scope, and so are obviously stylised outputs like animation. You switch it on in the post settings before publishing, and viewers then see a line on the video saying the creator disclosed it as AI-generated. Two things matter before you tap it. TikTok says the label does not affect distribution as long as the video follows its guidelines, and you cannot remove it once the post is live. Content made with TikTok's own effects is labelled automatically, and TikTok's support page says it may also apply the label itself to content it identifies as fully generated or significantly edited with AI.
LinkedIn and Content Credentials: the label that arrives without you
LinkedIn is the outlier. There is no AI toggle at upload and no published requirement to declare that you used AI. What it does instead is show a small Content Credentials icon on images and video that arrive cryptographically signed under the C2PA standard — the same provenance metadata Meta reads and TikTok's automatic labelling leans on. Tap it and you can see whether AI was used to generate or modify the file, which app it came from, and who issued the credential. LinkedIn's own help page is candid about the limits: it says it is not yet possible to identify and label all AI-generated content, and that what you see depends on assertions made by whoever signed it. Provenance is also fragile, since re-encoding or screen recording can strip it. A missing credential proves nothing either way.
The EU AI Act now sits on top of all of this
If part of your audience is in the EU, there is a legal layer above the platform rules as of this month. Article 50 of the EU AI Act applies from 2 August 2026. It requires anyone deploying an AI system to create a deepfake — generated or manipulated image, audio or video resembling real people, objects, places or events that would falsely appear authentic — to disclose that it is artificially generated, clearly, perceivably, and at the latest on first exposure. Evidently artistic, satirical or fictional work still gets disclosed, but in a way that does not spoil the piece. Usefully for creators, the European Commission's guidance says the marking obligation does not bite where the AI performs an assistive function for standard editing. That is the same line the platforms drew.
What this means for a normal clipping workflow
For most people cutting long video into clips, nothing they do needs a label. Transcribing, trimming, reframing to vertical, generating captions, ranking hooks, drafting titles — every platform above puts that on the assisted-editing side, and the EU guidance treats standard editing help the same way. The obligation starts when you generate or alter footage of a person or an event in a way a viewer could take as real: a synthetic presenter, a cloned voice that is not yours, a shot of a real place that never existed. Build one check into your process. Before a clip goes out, ask whether any frame or second of audio depicts something that did not happen. If yes, disclose everywhere, not just on the strictest platform. Tools like FrameOS sit firmly on the assisted side, because they cut and caption footage you actually recorded.
FAQ
Do I need to label AI-generated captions or subtitles?
No. YouTube's help page lists caption creation among the minor edits that do not require disclosure, and its 2024 announcement grouped automatic captions with productivity uses like scripts and content ideas. TikTok and Meta both aim their rules at realistic depictions of people and events, not at editing assistance. Add captions freely, on every platform.
Does AI reframing or auto-cropping count as altered content?
No. YouTube's test is whether an edit is primarily aesthetic and leaves the viewer with an accurate impression of what actually happened. Reframing footage you recorded changes the framing, not the event. TikTok explicitly puts cropping among the small edits that need no label. The rules are about depiction, not about which tools touched the file.
What happens if I do not disclose AI content on YouTube?
YouTube's help page says creators who consistently choose not to disclose may have a label applied manually, and may face penalties including removal of the content or suspension from the YouTube Partner Program. Since May 2026 YouTube has also applied labels on its own when its systems detect significant photorealistic AI use without a disclosure.
Does adding an AI label reduce your reach?
The platforms say no. TikTok states that turning on the AI-generated content setting does not affect distribution provided the video follows its guidelines, and YouTube said in May 2026 that disclosure does not affect monetisation eligibility or recommendations. Meta is the exception worth watching: it reserves the right to apply a more prominent label to content it judges high-risk.
Sources
- Disclosing use of altered or synthetic content · YouTube Help
- Improving AI labels for viewers and creators · YouTube Blog
- Our Approach to Labeling AI-Generated Content and Manipulated Media · Meta
- AI-generated content · TikTok Support
- Content credentials · LinkedIn Help
- Transparency obligations under Article 50 of the AI Act · European Commission
Related pages
Keep reading
More AI →The best AI tools for social media, by the job you need done
Most AI tool lists are directories. This one is organised by job, because the useful question is not which tool is best overall, it is which step of your week you want to stop doing by hand.
9 minute readThe best AI transcription tools, judged on measured accuracy
On clean read speech almost every engine looks excellent. On meetings, accents and crosstalk the same models lose several points of accuracy — and the published numbers show exactly where.
8 minute readThe best AI writing tools for creators, and what each one is actually for
Most roundups list forty tools that all do the same thing. The useful split is between general assistants, the editing layer, and tools that write from material you already have.
8 minute readTurn one long video into a week of posts
300 credits for 3 days · no card.
Start free