Current FTC and platform-by-platform AI content disclosure rules in 2026. TikTok, Meta, YouTube, X, LinkedIn. What requires disclosure, what does not, and the soft-flags where rules have shifted recently.
Last verified 2026-05-22
Direct answer: In 2026 the FTC and every major social platform have AI-content disclosure rules that have shifted multiple times since 2023. Broadly: synthetic media depicting real people, AI-generated content of events that did not happen, and AI-altered political content carry the strongest disclosure obligations. Pure AI-written text (captions, articles) generally does not require disclosure on social platforms but does in commercial-endorsement contexts under FTC rules. This page summarizes current platform language as of 2026-05-22; verify on each platform before relying on a specific phrasing — rules have changed several times across 2024-2026 and will continue to.
AI content disclosure is the most fast-moving regulatory area in creator content in 2026. The FTC updated endorsement guides to address AI-generated reviews and synthetic-person endorsements. The EU AI Act came into force with specific disclosure obligations for synthetic content. TikTok, Meta, YouTube, X, and LinkedIn all updated their AI-content labeling rules multiple times between 2024 and 2026. Multiple US states added political-deepfake statutes. The space is genuinely confusing and any single-source guide has a shelf life of months.
This page is the working summary as of 2026-05-22. We have softened claims where rules have shifted recently or where wording is platform-specific; soft-flags appear inline. We have NOT paraphrased legal language in ways that change its meaning; where the exact platform or regulator wording matters, we point you to the official source. Kompozy is not your lawyer and this page is not legal advice — for high-stakes commercial content, verify on each platform's current policy page and consult counsel where the stakes warrant it.
The overall pattern across 2026: the regulatory direction is unambiguously toward more disclosure, more granular labeling, and stricter enforcement on synthetic media of real people. Do not bet against that direction. When in doubt, disclose.
The US Federal Trade Commission updates its endorsement and testimonials guidance regularly. As of 2026, the working rule is that endorsements must be honest, not misleading, and properly disclose material connections. AI-specific updates across 2023-2026 have addressed: fake reviews (including AI-generated ones), synthetic endorsements (AI-generated person endorsing a product is a misleading endorsement unless properly disclosed), and undisclosed paid relationships in AI-generated content.
Practically: if you are using an AI avatar or AI voice in commercial content (ad, sponsored post, product endorsement), the FTC stance is that the synthetic nature should be clear to the average viewer. The exact required phrasing has not been codified in the same way as #ad disclosure, but the trend across 2024-2026 enforcement actions has been toward requiring obvious labeling. Soft-flag: verify current FTC.gov endorsement guidance for the latest language.
TikTok's current rule (as of 2026-05-22): creators must label AI-generated content that shows realistic-looking scenes, including realistic people, places, or events that did not happen. The platform provides an "AI-generated" label that creators can toggle on at upload. TikTok's platform also automatically detects and labels some AI-generated content using Content Credentials and provider partnerships.
Categories typically requiring the label: AI avatar video of a real or realistic-looking person, AI-altered video of a real person, AI-generated scenes depicting events that did not happen, voice-cloned narration of a real person. Categories typically not requiring the label: AI-assisted writing (captions, titles), AI-generated music backgrounds (separate music rules apply), text-only AI content.
Soft-flag: TikTok has updated this rule at least twice across 2024-2026; verify on creators.tiktok.com for current language.
Meta introduced "Made with AI" labels in 2024 and has revised the labeling approach multiple times since. As of 2026, the current rule is that content where AI created or significantly altered a depiction of a person (including AI-generated images of real-looking people, AI-altered photos or videos, AI-generated voice of a person) should carry the Made with AI label. Meta's detection systems automatically apply some labels using Content Credentials; creators are also expected to self-disclose when uploading AI-generated content.
Categories typically requiring the label: AI avatar video, AI-cloned voice of a person, AI-generated images depicting real-looking people, significant AI alterations to photos or video of real people. Categories typically not requiring: minor AI editing assistance (face-smoothing filters, basic touch-ups), AI-assisted writing in captions, AI-generated abstract art or non-person imagery.
Soft-flag: Meta's labeling approach has been refined repeatedly since the initial 2024 rollout in response to creator feedback (initial implementation over-labeled minor edits); verify on transparency.meta.com for current language.
YouTube's altered-or-synthetic-content rule (introduced 2024, refined across 2024-2026) requires creators to disclose when uploaded content contains realistic altered or synthetic material — content that could be mistaken for real people, events, or places. At upload, creators check an "altered or synthetic content" disclosure; YouTube displays a label on the video and in the description.
Categories typically requiring disclosure: AI-generated video of a real-looking person, AI voice-clone narration of a real person, AI altered footage of real events, deepfake-style content. Categories typically not requiring: animations and clearly non-realistic content, beauty filters and color correction, background blur, AI-generated B-roll that is clearly not depicting real events.
Some categories trigger additional restrictions beyond labeling — particularly content related to elections, health crises, or public figures, where YouTube has stricter enforcement and may remove content regardless of disclosure. Verify on support.google.com/youtube for current rule language.
X (formerly Twitter) has a synthetic and manipulated media policy that has been in place in some form since 2020 and was updated in 2023-2024 to address AI content specifically. Content that is manipulated in deceptive ways or could cause confusion or harm carries labeling or restrictions. Community Notes (crowdsourced fact-checks) often flag synthetic media regardless of creator self-disclosure.
There is no general "this is AI" label requirement on X for non-deceptive AI content. Self-disclosure is voluntary. Soft-flag: X's moderation approach has shifted multiple times since 2022; verify on help.x.com for current language.
LinkedIn's AI-content policies are more permissive than the social-video platforms, reflecting the platform's different content mix. AI-generated articles and posts are allowed and not specifically labeled. Synthetic media depicting real people (deepfakes) is restricted under the broader fake-content policy. LinkedIn does not have an analog to TikTok's or Meta's explicit AI label as of 2026-05-22.
The EU AI Act, which came into force across 2024-2026 with phased provisions, requires providers and deployers of AI systems generating synthetic content (text, image, audio, video) to mark output as artificially generated or manipulated, in machine-readable format where technically feasible. Content depicting real people or events in deepfake style carries additional disclosure obligations toward the public. Member-state-level implementation is still in flux as of 2026-05-22.
For creators publishing to EU audiences, the practical impact is similar to platform-level requirements: synthetic media of real people requires disclosure; AI-assisted writing is treated more leniently; deepfake-style content carries the strictest obligations. Soft-flag: EU AI Act implementation is incomplete; consult current EU and member-state guidance for specifics.
Multiple US states across 2023-2026 have passed deepfake-specific statutes covering political deepfakes (Texas, California, Washington, and others), non-consensual sexual deepfakes (most US states by 2026), and right-of-publicity expansions covering voice and likeness cloning. Penalties range from civil damages to criminal charges. State-by-state variation is significant; check state law for your jurisdiction and the jurisdictions you publish to.
Kompozy does not bypass platform disclosure requirements. Personas (HeyGen avatar + ElevenLabs voice) are explicitly synthetic, and any platform publishing of persona-based video should be paired with the appropriate platform disclosure label at upload. The product surfaces a soft reminder on persona-format scheduling. For commercial endorsement content, you are responsible for the FTC-required disclosure. Kompozy provides the production tooling; legal compliance and disclosure remain creator-side. Pricing: Founding $39/mo BYO (signups close 2026-08-31), Creator $49/mo / 2,500cr, Starter $99/mo / 5,500cr, Pro $299/mo / 18,000cr, Agency $799/mo / 55,000cr.
On TikTok, Meta, and YouTube: yes for AI content depicting realistic people or events. On X and LinkedIn: lighter rules but the broader synthetic-media policies apply. Under FTC rules: yes in commercial endorsement contexts. Specific platform language has shifted multiple times since 2023; verify current rules before relying on a specific phrasing.
AI-generated endorsements (synthetic-person endorsing a product) are subject to general FTC endorsement rules — they must be honest, not misleading, and disclose material connections. The trend across 2024-2026 enforcement has been toward requiring obvious labeling of synthetic endorsements.
Generally no on social platforms. AI-assisted captions, articles, scripts, and emails do not currently require platform-level labeling on TikTok, Meta, YouTube, X, or LinkedIn. Exception: AI-generated reviews and endorsements in commercial contexts under FTC rules.
TikTok provides an "AI-generated" content label that creators toggle at upload. It is required for AI-generated content showing realistic-looking people, places, or events that did not happen. Some content is also auto-labeled via Content Credentials integrations.
A label Meta applies to content where AI created or significantly altered a depiction of a person. Creators can self-disclose at upload; Meta also auto-labels via Content Credentials. The labeling approach has been refined repeatedly since the 2024 rollout.
A creator-level disclosure at upload for "altered or synthetic content" — realistic content that could be mistaken for real people, events, or places. The label displays on the video and in the description.
Cloning your own voice with platform-required verification is legal in most jurisdictions. Cloning someone else's voice without explicit written consent exposes you to right-of-publicity, defamation, and in many jurisdictions specific deepfake or voice-cloning statutes. The technology being available does not make the use legal.
The trend across 2023-2026 has been unambiguously toward more disclosure, more granular labeling, and stricter enforcement on synthetic media of real people. Do not bet against that direction. When in doubt, disclose.