Platform disclosure rules for realistic AI-generated video are enforced policies, not optional guidance: undisclosed realistic AI content can be labeled for you or removed. Regulators are adding their own rules on top, most notably the EU AI Act's deepfake labeling duty. If you're producing AI video at any volume, you need to know what each platform actually requires, what the law does and doesn't say, and how to disclose cleanly.
This guide is general information, not legal advice. Platform policies and regulations change; check the linked primary sources and get legal advice for your situation.
Why AI Video Disclosure Became Non-Negotiable in 2026#
Three forces converged to make AI content disclosure mandatory rather than aspirational.
Platform trust and misinformation risk. The major platforms — YouTube, TikTok, Meta — introduced explicit policies asking creators to label AI-generated or AI-altered content, particularly anything that depicts real people, real events, or anything a reasonable viewer could mistake for genuine footage. On TikTok and YouTube, failing to label realistic AI content is a policy violation, not just an ethical shortcoming.
Regulation. The EU AI Act's transparency rules (Article 50) apply from 2 August 2026 and include a duty for businesses that publish deepfakes to label them. In the US, the FTC's rules against fake reviews and testimonials apply whether a fake testimonial is written by a person or generated by AI.
Audience trust. Formats that sell on looking real, like testimonials and raw UGC, depend on the viewer believing what they see. If a viewer later finds out it was AI and nobody said so, it can feel like deception even when none was intended. Disclosing upfront avoids that.
The result is an environment where compliance and trust-building are the same action. Understanding the rules per platform is the starting point.
YouTube's Synthetic Media Disclosure Requirements#
YouTube's altered or synthetic content policy, introduced in 2024, requires disclosure for realistic content that:
- Depicts a real person saying or doing something they never said or did
- Shows realistic footage of a real place or event that never occurred
- Has been meaningfully altered in a way a viewer could mistake for genuine footage
The disclosure mechanism is an "altered or synthetic content" setting in the YouTube Studio upload flow. When you turn it on, YouTube adds a label to the video's expanded description, and for videos on sensitive topics (such as health, news, elections or finance) it shows a more prominent label on the video player itself. YouTube can also add a label itself when creators don't, especially where the content could mislead.
What YouTube does not require disclosure for:
- AI tools used for editing assistance (auto-captions, color grading, background removal) where the footage is otherwise genuine
- Animation or clearly fantastical AI-generated visuals that no viewer would mistake for real
- AI-generated voiceover when the creator's own voice isn't being replicated
Where creators make mistakes: a common gap is UGC-style content — AI avatars made to look like real people giving authentic testimonials. The authentic AI UGC format is a popular ad style, but a realistic AI person presented as real is exactly what the disclosure setting exists for. Creators who consistently skip disclosure on realistic content risk penalties from YouTube, including content removal or suspension from the YouTube Partner Program.
Practical disclosure in your upload flow: build the YouTube synthetic media toggle into your upload checklist. If your production workflow involves AI video generation at scale, create a standard operating procedure that includes the disclosure step before publish — not as an afterthought but as part of the final QA check on every piece.
TikTok's AI Content Label Policy#
TikTok's AI-generated content policy separates what it requires from what it encourages:
Required: label realistic AI content. TikTok's Community Guidelines require creators to label AI-generated content that contains realistic images, audio or video. TikTok may remove content that violates its guidelines, including unlabeled realistic AI content.
Encouraged: label anything fully generated or significantly edited by AI. That includes videos where the main subject is shown doing or saying something they didn't, or where their appearance is substantially changed, for example with a face swap.
How to label: turn on the "AI-generated content" setting under More options before you post. The post then shows a "Creator labeled as AI-generated" label. You can also disclose in text, a sticker or the caption.
Auto labels: TikTok may apply an "AI-generated" label itself when you use TikTok's AI effects or upload content carrying C2PA Content Credentials. Once an auto label is applied, you can't remove it.
Labeling doesn't cost reach. TikTok states that turning on the AI-generated content setting won't affect the distribution of your video as long as it doesn't violate its Community Guidelines. So there's no reach reason to skip the label.
Banned even with a label: AI content showing fake authoritative sources or crisis events, public figures falsely shown endorsing something, anyone under 18, or adult private individuals without their permission.
Ad creative: TikTok's advertising policies have their own rules for AI-generated ads. Check the current ad policies before running AI creative, since an organic label doesn't carry over to ad review.
Instagram and Meta's Disclosure Approach#
Meta's disclosure framework spans Instagram, Facebook, Threads, and Meta's ad network. It differs from YouTube and TikTok in one key way: Meta relies more heavily on automated detection and labels content proactively, rather than requiring creator opt-in alone.
For organic content, Meta's AI detection system adds an "AI Info" label to posts it identifies as AI-generated or AI-altered. This label is surfaced in the post's context menu rather than directly overlaid on the content — less prominent than TikTok's approach, but present and accessible to viewers who look.
What Meta explicitly requires from creators:
- For organic posts with photorealistic video or realistic-sounding audio that was digitally created or altered, Meta asks creators to use its AI label in the post composer (shown to viewers as "AI info"), and may apply penalties if you don't
- For advertising, Meta's ad review policy requires the AI content flag for any ad that uses generative AI to depict people, places, or events in a realistic way
Practical caveat: Meta's automated detection has a meaningful false-positive rate for high-quality AI content. Proactively labeling your AI-generated content — rather than waiting for Meta's system to do it — prevents the scenario where your post gets a system-generated label that wasn't placed by you, which can read as a flag rather than a disclosure.
For brand awareness campaigns running across the Meta ad network, the brand safety consideration is real. Major brands monitoring ad placement may exclude content that shows auto-generated AI labels without creator disclosure context. Getting ahead of it with proactive labeling is both a compliance step and a brand safety step.
The Regulatory Landscape: What's Coming#
Platform policies sit alongside a legal framework that is still taking shape. Here is what the main rules actually say.
EU AI Act, Article 50 (transparency): Article 50 applies from 2 August 2026. It does not require a label on every realistic AI video that reaches an EU audience. It puts specific duties on specific parties:
- Providers of generative AI systems (the companies that build the tools) must mark AI output in a machine-readable way. Systems already on the market before 2 August 2026 have until 2 December 2026 for this marking duty.
- Deployers (people or businesses using an AI system in a professional capacity) must clearly disclose deepfakes they publish. Purely personal, non-professional use falls outside the Act. The Commission's FAQ says that using AI in a business, trade or freelance activity generally makes you a deployer, and that a company stays the deployer when staff or contractors make the content for it.
- A deepfake is AI-generated or manipulated image, audio or video that resembles real people, objects, places, entities or events and would falsely appear authentic. The Commission's Article 50 FAQ says all of these conditions must be met, judged in context: if viewers wouldn't expect the content to be authentic, it may not count.
- For evidently artistic, creative, satirical or fictional work, the duty is lighter: disclose the deepfake in a way that doesn't spoil the work.
- Where the duty applies, disclosure must be clear, visible or audible, and given at the viewer's first exposure. A tool's machine-readable watermark alone doesn't satisfy it.
Whether a given video is a deepfake depends on these criteria and its context, so there's no one-line rule. If you publish realistic AI footage that viewers could take for a real person or real event, check it against the FAQ's criteria, and get legal advice if you're unsure.
FTC (US): The FTC revised its Endorsement Guides in 2023. In 2024 it adopted the Consumer Reviews and Testimonials Rule, effective 21 October 2024, which bans fake or false consumer reviews and testimonials, including ones generated by AI, along with buying sentiment-based reviews and similar practices. There is no blanket FTC rule that every AI spokesperson must be labeled. The FTC says an AI stock avatar isn't a consumer review, but it can be a testimonial; it's a problem when the testimonial is fake or false, for example an AI "customer" describing an experience no real customer had, or a celebrity's likeness used without permission. Separately, any ad that misleads consumers, including through an AI avatar, can still be deceptive under the FTC Act. Two practical points follow. A label does not make a fake testimonial lawful: an AI "customer" describing results no real customer had is the problem itself, so don't publish it, labeled or not. And where an AI spokesperson is legitimate (a presenter reading your own claims, for example), disclosing that it's AI helps avoid misleading viewers into thinking a real person is vouching for you.
US states: Several states have passed laws on AI-generated content, mostly aimed at election deepfakes and at using someone's likeness or voice without consent. Scope, status and court challenges vary by state, so check the rules where your audience is.
What this means for your workflow: regulation adds a legal dimension to what used to be only a platform policy question. For brands and agencies, it's worth getting a legal review of your AI content disclosure practices for both the platforms you use and the places your audience lives.
How to Disclose Without Hurting Your Content#
The practical worry for most creators and brands is whether a label hurts performance. Here's what's actually known:
- Distribution: TikTok states its AI-generated content setting doesn't affect a video's distribution as long as the video follows its Community Guidelines. Labeling is not a reach penalty there.
- Audience response: how viewers react to a label varies by content and audience. There's no reliable public benchmark, so if it matters to you, test it on your own account.
- Authenticity formats: content that sells on looking real — testimonials, raw-feeling UGC, hidden-camera style footage — is where an undisclosed AI origin does the most damage when it comes out, and where platform and legal rules bite hardest. If the format relies on looking "real," disclose.
- Your label vs. theirs: labeling yourself means the disclosure is in your words and on your terms, rather than a platform label appearing after the fact.
The best disclosure language: "Created with AI tools" is neutral and accurate. "AI-generated" is the standard platform language. "Built using AI video generation" for description or caption context. Keep it factual and brief — the disclosure doesn't need a paragraph of explanation.
For social proof content, keep two things apart. A customer testimonial must reflect a real customer's real experience; an AI-invented customer story is a fake testimonial under the FTC rule, and a label doesn't fix that. When you use AI to present real customer feedback, for example an AI presenter reading a genuine review, disclose the AI and don't treat the label as a performance risk to minimize.
Building Disclosure Into Your AI Video Production Workflow#
Compliance at scale requires treating disclosure as a production step, not a post-publishing afterthought. The operations that handle this best have three things in place.
A content classification system that tags every video asset at creation time: fully AI-generated, AI-assisted (AI elements in otherwise live footage), or native (no AI generation in the final output). The classification determines which disclosure labels apply. Without this tagging at creation, disclosure decisions happen under time pressure at upload — when they're most likely to be skipped.
A platform-specific publish checklist that maps each content type to its disclosure requirements per platform. Example format: a simple spreadsheet or project management template with one row per piece of content, columns for each platform, and checkboxes for disclosure steps completed before publish. For teams managing AI video content calendars across multiple platforms and multiple creators, this checklist becomes the shared compliance layer.
A regular audit cadence — monthly or quarterly — that checks published content against disclosure requirements and reviews any enforcement actions. Platforms update their policies; regulatory requirements change, and the EU's Article 50 guidance is still new. An audit cadence ensures your workflow stays current rather than drifting out of compliance as rules evolve.
The goal is to make disclosure automatic, not effortful. When it's baked into production — asset tagged at creation, checklist run at publish — it costs almost nothing per piece of content and eliminates the decision fatigue that leads to omissions.
If you're producing AI video at scale and want a workflow that makes compliance a built-in step rather than an afterthought, Mango's production system is designed to handle the volume that makes manual disclosure management impractical.
