You made a video with an AI voice, AI-picked B-roll, maybe one AI-generated shot — and now the upload screen is asking whether the video contains AI content. Say yes and you worry the algorithm buries it. Say no and you worry about a strike. Every platform words the question differently, the rules changed several times between 2023 and 2026, and the EU just layered a law on top. Most creators are guessing.
The direct answer: every major platform now requires you to disclose realistic AI-generated or AI-altered video, audio, and images. TikTok uses an AI-generated content label (plus automatic labeling via C2PA Content Credentials), Instagram and Facebook use the “AI info” label (formerly “Made with AI”) and require self-disclosure for photorealistic video or realistic-sounding audio, YouTube asks for an altered or synthetic content disclosure in Studio, and Pinterest applies an “AI modified” label. Labels are not required for clearly unrealistic or lightly assisted content — AI-written scripts, obvious narration over stock footage, captions, color correction, beauty filters — and no platform says a label reduces distribution. What gets removed or penalized is realistic AI content you didn't disclose.
This guide walks through each platform's rule, what does and doesn't need a label, whether labeling hurts reach, what the EU AI Act changes from August 2026, and how to keep disclosure consistent when you publish dozens of faceless videos a week through a pipeline like Vidpal.
Why do platforms label AI content now?
Platforms label AI content because realistic synthetic media became cheap and convincing at the exact moment their feeds filled with it. Between 2023 and 2025, tools like Sora, Veo, Runway, and ElevenLabs put photorealistic video and cloned voices in the hands of anyone with a browser. TikTok, Meta, and YouTube concluded that viewers needed a signal to separate recorded footage from generated footage, especially around news, elections, health, and finance. The label is that signal.
The second reason is liability. Regulators in the EU, the UK, and several US states pushed platforms to deal with deepfakes, and a visible label paired with a machine-readable marker is the cleanest way to show they did. TikTok joined the Adobe-led Content Authenticity Initiative in 2024, Meta built detection for invisible markers from Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock, and YouTube added automatic AI detection. The whole industry moved in the same direction within roughly 18 months.
For creators, the practical consequence is simple: disclosure went from courtesy to platform rule. It also changed how you should choose tools. Our Sora 2 vs Veo 3 comparison and best AI video generator roundup both flag which generators embed provenance metadata, because that decides whether your uploads get labeled automatically before you touch a toggle.
What are Content Credentials (C2PA) and why do they matter?
Content Credentials are a tamper-evident metadata standard from the Coalition for Content Provenance and Authenticity (C2PA) that records how a piece of media was made — which tool generated or edited it and whether AI was involved — and travels inside the file. Think of it as a nutrition label baked into the video or image itself. Platforms read it on upload and can apply an AI label without you doing anything.
This matters because more generators attach Content Credentials by default every year. TikTok said in May 2024 that it would automatically label uploaded images and videos carrying Content Credentials and would attach the credentials to AI content made inside TikTok so the label survives download in a newsroom post announcing it had joined the Content Authenticity Initiative. Meta's detection reads C2PA and IPTC markers. YouTube's help page says it may apply a label automatically based on C2PA metadata, its own AI tools, or its detection systems.
Two consequences follow. You may get labeled even if you never self-disclosed, which is fine. And stripping metadata to dodge a label is a bad idea: platforms treat undisclosed realistic AI as a violation, and their classifiers increasingly catch it without any metadata.
Do you have to label AI content on TikTok?
Yes, when it looks real. TikTok's Community Guidelines require creators to label AI-generated content that shows realistic-looking scenes or people, and TikTok's AI-generated content help page explains the label and the toggle. Content that is obviously not real — a cartoon, a surreal animation, an AI narrator reading a listicle over stock clips — does not require the label, though TikTok encourages transparency either way. Realistic AI content that is not disclosed, whether by the label or by a clear caption, sticker, or watermark, is treated as a violation and can be removed.
How to toggle it: when you post, open the post settings screen — the same screen where you set who can view the video and whether comments are allowed — and switch on the AI-generated content toggle. A small label appears on the published video telling viewers the creator disclosed it as AI-generated. Set it before you hit Post; it is far easier than dealing with a removal and appeal afterward.
Automatic labeling: TikTok auto-labels content made with its own AI effects and tools, and since May 2024 it reads Content Credentials on uploaded images and videos and labels them on its own. In November 2025 TikTok said it would start adding invisible watermarks to AI content made with TikTok tools and to uploads that carry Content Credentials, so the label sticks even when metadata gets stripped by a download or re-upload.
The see-less-AI control: in that same late-2025 update, TikTok announced an AI-generated content setting inside Manage Topics that lets viewers dial down, or up, how much AI content the For You feed shows them. That is a personalization control, not a penalty, but it is worth understanding. A labeled video can reach fewer of the people who chose to see less AI — that is the audience effect of a label, and it is very different from algorithmic suppression. If you run a faceless TikTok account, our TikTok automation guide covers how to keep the content itself worth watching regardless.
What are Instagram and Facebook's AI label rules?
Instagram, Facebook, and Threads share one label called “AI info,” which Meta renamed from “Made with AI” in July 2024 after photographers complained that lightly retouched photos were being tagged as AI-made. Meta applies it two ways: automatically, when it detects industry-standard AI markers in a file, and manually, when you self-disclose at the point of posting.
The self-disclosure requirement is specific. In its February 2024 announcement, Meta said it would “require people to use this disclosure and label tool when they post organic content with a photorealistic video or realistic-sounding audio that was digitally created or altered, and we may apply penalties if they fail to do so.” Read the scope carefully: photorealistic video and realistic-sounding audio. An AI illustration in a carousel or a clearly synthetic narration over B-roll is not what that sentence describes, though you are free to label it anyway.
How to add the label on Instagram: when you share a Reel or a post, the final sharing screen has an Add AI label toggle. Instagram moves it between versions — if you can't see it, check Advanced settings at the bottom of the share screen. On Facebook the equivalent option sits in the post composer. Since late 2024, Meta shows the label more prominently on content it detects as fully AI-generated and tucks it into the post's menu for content that was only edited with AI tools.
On industry signals: Meta said it built tools to read invisible markers and metadata aligned with the C2PA and IPTC standards from images made with Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock tools. So an image from those generators may get the label even if you skipped the toggle. Meta also said at the time that detection for audio and video lagged because generators weren't embedding signals at scale — which is exactly why the self-disclosure rule targets video and audio. If strikes are your wider worry, how to avoid Instagram account suspension covers the other tripwires for automated accounts.
When does YouTube require the altered or synthetic content disclosure?
YouTube requires disclosure when a video contains realistic content — something a viewer could easily mistake for a real person, place, or event — that was meaningfully altered or generated with AI. The setting lives in YouTube Studio's upload flow. YouTube's help page currently describes it as the “AI use” question in the Attributes section, where you answer Yes if the content meets the bar; earlier versions of the page called it the “Altered content” checkbox. Viewers then see a label — it has read “Altered or synthetic content” — in the expanded description, and directly on the player for photorealistic content or sensitive topics.
YouTube's own examples of content that must be disclosed: making it appear that someone gave advice they never gave; showing a realistic tornado or other weather event moving toward a real city when it didn't happen; AI-generated realistic video of a match between two real professional tennis players; AI-generated extra footage of a real place, such as a surfer in Maui for a travel promo; making it look like a real person was arrested or admitted to something they didn't do; and, in the current version of the page, AI-generated music.
Examples YouTube says you don't have to disclose: someone riding a unicorn through a fantastical world; green screen used to show someone floating in space; fully animated videos; beauty, color, and lighting filters; production assistance such as using AI to write an outline, script, title, or thumbnail; caption creation; sharpening, upscaling, and audio repair; gameplay footage; and cloning your own voice for voiceovers or dubs. That last exemption matters a lot for narrated channels.
The consequence for ignoring it: YouTube says creators who consistently choose not to disclose may have a label applied manually or face penalties including content removal or suspension from the YouTube Partner Program. The label itself is not a monetization issue. Our sibling post on YouTube's AI content monetization policy covers what actually threatens revenue — mass-produced, repetitive content — and why a disclosed, well-made AI video is not in that category.
What about Pinterest and X?
Pinterest labels Pins it identifies as generated or altered with AI as “AI modified,” and the label appears at the bottom-left of the Pin close-up, per Pinterest's help center. Pinterest detects AI content three ways: IPTC metadata in the file, its own classifiers that catch AI content even when there are no obvious markers, and self-reports from the content owner. In 2025 Pinterest also began rolling out a “see fewer” option in a Pin's three-dot menu for generative AI Pins in categories like beauty and art, so viewers who are tired of AI imagery can tune it down.
X takes a narrower approach. Its synthetic and manipulated media policy is aimed at deceptive media that could cause harm — X may label, restrict, or remove it — and Community Notes often supply the context in practice. There is no creator-facing AI toggle on X at the time of writing, so the practical rule is to put your disclosure in the post text itself and check X's help center for the current wording.
What does NOT need an AI label?
Most of what a faceless creator produces does not need a label. Every platform's rule targets realism, not assistance. Here is the list, in roughly the order people ask about it.
AI-written scripts: exempt everywhere. YouTube lists scripts, outlines, titles, and thumbnails under production assistance, and no platform treats text help as synthetic media. The viewer is watching a video, not reading your drafting process.
AI voiceover on obviously narrated faceless videos: a synthetic narrator reading a script over stock footage, screen recordings, or slides is not a realistic depiction of a real person saying something they didn't. YouTube explicitly exempts cloning your own voice, and a generic text-to-speech voice isn't imitating anyone. Label it if the voice imitates a specific real person or the video pretends the narration is a genuine interview. Our realistic AI voiceover guide goes deeper, and the free text-to-speech tool is a quick way to hear what an obviously narrated voice sounds like.
AI-generated B-roll that is clearly illustrative: an abstract animation, a stylized illustration, a fantasy scene, a diagram. You only cross the line when the generated visual depicts a real place, a real event, or a real-looking person as if it were footage. A quick test: if a viewer might screenshot it and share it as a real photo, disclose. The free text-to-image tool is useful for making visuals that are deliberately stylized rather than photoreal.
Captions and subtitles: exempt on every platform. Auto-generated captions are the most common AI use in short-form and nobody labels them.
Thumbnails, titles, descriptions, and hashtags: exempt as production assistance. The exception is a photorealistic AI thumbnail of a real person doing something they didn't do — that is a realistic depiction and falls under the disclosure rules.
Color correction, lighting fixes, sharpening, upscaling, and audio repair: exempt. Platforms count these as standard editing, and the EU AI Act's marking rule carves out standard editing functions too.
Beauty filters and background blur: exempt. YouTube names both directly.
Animations and fully animated videos: exempt as long as they are obviously animated. A cartoon explainer does not need a label; a photoreal animation of a real newsroom does.
When in doubt, label. No platform penalizes over-disclosure, and a label on a borderline video costs you nothing compared to a removal.
Does labeling AI content hurt reach?
No platform has said an AI label reduces distribution, and none of the published policies list the label as a ranking signal. TikTok, Meta, and YouTube all frame labels as viewer context, and they apply them automatically to plenty of content whose reach they have no reason to cut. The reach risk runs the other way: undisclosed realistic AI content can be removed outright, repeat offenders face account-level penalties, and a removed video reaches nobody.
What does hurt reach is worth separating from the label. First, viewer controls: TikTok's Manage Topics setting and Pinterest's see-fewer option let people opt out of AI content, so a labeled post reaches fewer of those specific viewers. Second, repetitive low-value output: platforms have spent 2025 and 2026 tightening rules on mass-produced, templated content, and that is enforced on the content, labeled or not. Third, trust: an audience that discovers undisclosed AI in what they thought was real footage stops watching, and no algorithm fixes that.
You will find creators online claiming a label tanked their views and creators claiming it changed nothing. We have not seen credible evidence of a label penalty, and the platforms' own statements point away from one. If reach dropped after you started labeling, look at the content first — hook, pacing, topic fatigue — before blaming a disclosure toggle. For the monetization side, where the real 2026 rule changes happened, read YouTube's AI content monetization policy.
What does the EU AI Act change in August 2026?
The transparency obligations in Article 50 of the EU AI Act applied from 2 August 2026 as scheduled. The EU's 2026 AI Omnibus (Regulation (EU) 2026/1744, in force from 27 July 2026) postponed other parts of the Act but left Article 50 untouched, apart from a grace period to 2 December 2026 for providers of generative systems already on the market to meet the machine-readable marking duty. Two of the obligations matter for creators. First, providers of AI systems that generate synthetic audio, image, video, or text must ensure the outputs are marked in a machine-readable format as artificially generated or manipulated — that is the tool makers' job, and it is why Content Credentials and invisible watermarks are spreading so fast. Second, deployers — which includes you, if you publish — who use an AI system to generate or manipulate a deepfake must disclose that the content has been artificially generated or manipulated.
A deepfake under the Act is AI-generated or manipulated image, audio, or video that resembles real persons, objects, places, entities, or events and would falsely appear authentic or truthful. That is close to the platforms' realism test, which is convenient: if a video passes TikTok's or YouTube's bar for disclosure, it almost certainly passes the EU's. The Act carves out evidently artistic, creative, satirical, or fictional work, where you only need to disclose in a way that doesn't spoil the piece, and AI-generated text on matters of public interest is exempt if it went through human editorial review with someone taking responsibility for it.
Who it reaches: deployers in the EU and, broadly, anyone whose AI output is used in the EU, which for a global creator effectively means everyone with European viewers. Expect enforcement to run mostly through the platforms' own labels and metadata rather than regulators reviewing individual TikToks. The European Commission has been developing a code of practice on marking and labeling AI-generated content and has discussed transition periods for some obligations, so check the current text before treating any one date as final. The practical playbook doesn't change: label realistic content, keep a disclosure sentence in your caption, and never strip provenance metadata.
AI label rules at a glance, platform by platform
TikTok: label required for AI content showing realistic-looking scenes or people. Toggle the AI-generated content switch in post settings. TikTok auto-labels content made with its AI tools and uploads carrying Content Credentials, and adds invisible watermarks. Viewers can reduce AI content via Manage Topics. Unlabeled realistic AI content can be removed.
Instagram: “AI info” label. Self-disclosure required for photorealistic video or realistic-sounding audio that was digitally created or altered. Use the Add AI label toggle on the share screen. Meta may apply the label automatically from C2PA and IPTC markers and may apply penalties for non-disclosure.
Facebook: identical to Instagram — same “AI info” label, same self-disclosure rule, same automatic detection. The toggle lives in the post composer. Threads follows the same policy.
YouTube: answer the AI use question in Studio (Yes if the video contains realistic altered or synthetic content). Label appears in the expanded description, and on the player for photorealistic or sensitive-topic videos. YouTube may label automatically. Repeatedly skipping disclosure can lead to removal or suspension from the Partner Program.
Pinterest: “AI modified” label applied from IPTC metadata, Pinterest's classifiers, or self-report, shown at the bottom-left of the Pin close-up. Viewers get a see-fewer option for generative AI Pins in some categories.
X: no creator-facing toggle. Deceptive manipulated media may be labeled, restricted, or removed under the synthetic and manipulated media policy. Put the disclosure in your post text.
EU (from 2 August 2026): tool makers must mark AI output machine-readably (providers of systems already on the market before that date have until 2 December 2026); anyone publishing a deepfake must disclose it. Artistic, satirical, and fictional work gets a lighter disclosure duty.
How to handle AI labels when you auto-publish at volume
Decide your labeling policy once, at the channel or template level, not per post. If you hand-publish three videos a week, flipping a toggle on each platform is manageable. At twenty videos a week across five platforms it is a hundred toggles, and inconsistency creeps in — one video labeled on TikTok but not on Reels, another with the YouTube box ticked by accident. Labels also don't travel between platforms: cross-posting a labeled TikTok to Reels does not carry the disclosure, so each destination needs its own. Our guide to cross-posting to Instagram, YouTube, TikTok, and Facebook covers what else gets lost in transit.
Where Vidpal fits: Vidpal produces narrated faceless content — an AI voiceover over stock footage and B-roll, with word-level captions, rendered 9:16, and published on a schedule. Under every platform's realism test, most of that output does not require a label: an obvious narrator over illustrative visuals is production assistance, not a synthetic depiction of a real event. Our recommendation is stricter than the minimum. Disclose whenever the visuals could be mistaken for real footage — AI-generated B-roll of a real place, a real-looking person, a realistic reconstruction of an event — and keep a standing disclosure sentence in your caption template, something like “Narrated with AI; some visuals are AI-generated.” That one sentence satisfies TikTok's caption-or-sticker route, gives Meta and YouTube reviewers context if a classifier flags the video, and is the deployer disclosure the EU AI Act asks for. The free Instagram caption generator is a quick way to draft a template that carries it naturally.
The workflow that scales: classify each content template as illustrative or realistic when you set it up, not video by video. Bake the disclosure sentence into every caption template. For realistic templates, switch the platform toggles on as part of the publishing step and leave them on. Use your review queue for spot checks rather than per-post decisions — Vidpal's approve-or-auto-publish queue is where you catch the occasional generated shot that looks too real. Never strip metadata to avoid a label. And when you compare tools, prefer ones whose generators embed Content Credentials, because an auto-applied label is a label you never have to think about. The step-by-step production side is covered in how to make faceless videos with AI and how to make Reels without showing your face.
Frequently Asked Questions
Do you have to label AI content on TikTok? Yes, if it shows realistic-looking scenes or people. TikTok's Community Guidelines require the AI-generated content label, or a clear caption, sticker, or watermark, on realistic AI content, and unlabeled realistic content can be removed. Clearly unrealistic content — animation, surreal visuals, an obvious AI narrator over stock clips — doesn't require it, though TikTok encourages disclosure. Turn the toggle on in post settings before you publish.
How do I label AI-generated content on Instagram? On the final share screen for a Reel or post, switch on the Add AI label toggle; if it isn't visible, look under Advanced settings. Instagram then shows the “AI info” label on the post. Meta requires this self-disclosure for photorealistic video or realistic-sounding audio that was digitally created or altered, and it may also apply the label automatically when it detects AI markers in the file.
Does labeling AI content hurt reach? Not according to any platform policy. TikTok, Meta, and YouTube describe labels as viewer context, not a ranking factor, and we have not seen credible evidence of a label penalty. The reach risks are undisclosed realistic content, which can be removed, and repetitive low-value content, which platforms suppress whether or not it is labeled. Some viewers use see-less-AI controls, so a labeled video may reach fewer of those specific people.
Do I need to label an AI voiceover on YouTube? Usually not. YouTube's disclosure rule targets realistic content, and its own examples of content that doesn't need disclosure include cloning your own voice for voiceovers and dubs, production assistance like scripts, and caption creation. A generic text-to-speech narrator over illustrative footage is not a realistic depiction of a real person. Disclose if the voice imitates a specific real person or the video presents the narration as a real recording.
What happens if I don't label AI content? On TikTok, undisclosed realistic AI content violates the Community Guidelines and can be removed. Meta has said it may apply penalties when people fail to disclose photorealistic video or realistic audio. YouTube says creators who consistently skip disclosure may have a label applied manually or face penalties including content removal or suspension from the YouTube Partner Program. Platforms also increasingly apply the label automatically, so skipping it rarely keeps a video unlabeled anyway.
Does the EU AI Act apply to me if I'm not in Europe? Broadly, yes, if your AI-generated content is used in the EU, which for most creators means having European viewers. From 2 August 2026, Article 50 requires anyone publishing a deepfake — realistic AI content resembling real people, places, or events — to disclose that it is artificially generated, with lighter duties for evidently artistic or satirical work. In practice, platform labels plus a disclosure sentence in your caption cover it. The 2026 AI Omnibus did not postpone Article 50; its only concession was a grace period to 2 December 2026 for providers of older generative systems on the marking duty, so the disclosure obligation for anyone publishing deepfakes applies now.
Disclose Clearly, Then Publish Everywhere
The rules across TikTok, Instagram, Facebook, YouTube, and Pinterest converge on one principle: if a viewer could mistake AI content for real footage, real audio, or a real person, label it; if it is obviously produced, assisted, or stylized, you don't have to. Labels are not a ranking penalty. Undisclosed realistic content is. Decide your policy once, keep a disclosure sentence in your captions, and let the platforms' automatic labeling do the rest.
If the bottleneck is doing this consistently across dozens of posts a week, that's what Vidpal is built for. It researches trending topics in your niche, writes short-form scripts in your brand voice, generates the AI voiceover, pulls relevant visuals and B-roll, burns in word-level animated captions, renders vertical 9:16 video, and auto-publishes to Instagram, TikTok, YouTube Shorts, Facebook, Pinterest, and X — with a review queue for spot checks and an analytics feedback loop that learns what performs. There's a free plan to start; see pricing for volume tiers, browse the use cases to see faceless channels running on it, and try the free tools for scripts, captions, voiceover, and visuals in the meantime.
