Do You Have to Label AI Videos on TikTok, YouTube and Instagram?

Almost everything in a faceless video is synthetic. The narrator is an AI voice, the scenes are AI images, and sometimes the music is generated too. So the same questions keep coming up: do I have to tag all of this as AI, does an AI voiceover count, and what happens if I don't?

None of the big platforms asks you to label every use of AI. They ask you to label content a viewer could mistake for something real. Where they disagree, and they do, is on the details that matter most to faceless channels, especially the AI voiceover. We read each platform's own help pages and policies on 18 October 2026, and we checked how our own posting code at Blipix handles labels. This is practical guidance, not legal advice. Every rule below links to the official page so you can check it, because these pages change.

In short (as of October 2026):

• TikTok wants a label on realistic-looking AI people or scenes. Generic text-to-speech narration and artistic styles like anime are exempt.
• YouTube wants disclosure for realistic depictions of real people, real events or places, scenes that never happened, and AI music as the main focus. AI-written scripts, thumbnails and a clone of your own voice don't count.
• Instagram and Facebook want a label on photorealistic video and realistic-sounding audio, and Meta's examples include "a reel narrated with a realistic AI-generated voiceover". Still images alone are exempt.
• Real people are the high-risk case everywhere. A cloned voice or realistic likeness of someone else needs a label on all three, and some uses stay banned even with one.
• Skipping a required label can get a video removed, restricted or labelled for you, and on YouTube repeated cases put Partner Program membership at risk.

The question every platform asks: could a viewer take it for real?

All three policies rest on the same idea. Using AI isn't the problem; misleading people about what they're watching is. So the test is realism, not the tool you used. A watercolor of a sailing ship is obviously made by someone. A photograph-style image of the same ship, dropped into a history video, invites viewers to believe they're looking at a real photo.

This matters more than most creators expect, because the default art style in Blipix (the one named Blipix) is photographic. Keep it for a history or true crime video and your frames sit on the realistic side of every platform's line. Pick an illustrated look from the style library, such as Watercolor, and the visuals stop being a labelling question on TikTok and YouTube. Instagram and Facebook are the exception, because of the voice.

TikTok, YouTube, Instagram and Facebook side by side

A summary of each platform's own pages as of October 2026. The sections below quote the exact lines.

TikTok: label anything that looks real

TikTok's Community Guidelines on edited media and AI-generated content (the version effective 24 September 2026) say creators must label "AI-generated or significantly edited content that shows realistic-looking scenes or people." The same section spells out what doesn't need disclosure, and for faceless creators it's the most useful list any platform publishes:

Disclosure isn't needed when: making small edits like color correction, reframing, or cropping; using artistic styles, like anime; using generic text-to-speech (TTS) narration, when the TTS isn't a recognizable voice of a known individual.

The cases that do need a label, even when nothing harmful is going on, include a swapped face, someone made to say something they didn't, and AI audio that mimics the voice of a real person.

How to switch the label on

TikTok's help page on AI-generated content gives the app steps: tap the Add post button, record or upload your video, tap Next, open More options (the three dots) and turn on the AI-generated content setting. The post then shows "Creator labeled as AI-generated". TikTok also accepts your own disclosure in the form of a clear caption, sticker or watermark, and the help page says the setting won't affect your video's distribution as long as it follows the Community Guidelines.

Auto-labels and enforcement

TikTok may add an "AI-generated" label by itself when you use its AI effects or upload a file carrying C2PA Content Credentials, the provenance metadata some AI tools attach to their output. Once TikTok applies an auto label, you can't remove it. Don't rely on detection to cover you, though: it only works when the file still carries that metadata.

Unlabelled realistic AI content "may be removed, restricted, or labeled by our team, depending on the harm it could cause." And some things stay off the platform even with a label: the likeness of private figures used without consent, fake news sources or crisis events, and public figures shown endorsing products or taking political positions they never took.

YouTube: disclose realistic content in Studio

YouTube's help page, now titled Disclosing use of GenAI content, asks creators to disclose content that:

• makes a real person appear to say or do something they didn't do
• alters footage of a real event or place
• generates a realistic scene that didn't actually occur
• creates music that's the main focus of the video

The music line is the one that catches people out. An AI song carrying a video needs disclosure on YouTube even when every visual is a cartoon.

The page is just as specific about what you can leave alone. Its "don't have to disclose" examples include clearly unrealistic content such as someone riding a unicorn through a fantastical world, beauty and lighting filters, background blur, using generative AI to create or improve an outline, script, thumbnail, title or infographic, caption creation, idea generation, gameplay footage from video games, and "cloning one's own voice to create voice overs or dubs".

Where the setting is

In YouTube Studio, on a computer or in the mobile app, go through the upload steps until you reach the Attributes section. Under "AI use", select Yes if your video meets the requirements, or No if it doesn't. YouTube then adds a label in the expanded description, and for photorealistic content or AI music it may also show one on the player. If you post Shorts, the rule is the same.

Auto-detection and what happens if you don't disclose

YouTube may apply the label itself for videos made with its own AI tools, files with C2PA metadata, and content its systems detect as AI. You get a notification in Studio and can usually correct a wrong call, except for labels triggered by YouTube's tools, C2PA metadata or a manual review. Creators who "consistently choose not to disclose" may get a label applied manually, have content removed, or be suspended from the YouTube Partner Program.

Instagram and Facebook: the voiceover rule

Meta's rule covers Facebook, Instagram and Threads, and for faceless creators it's the strictest of the three. According to Meta's help center, an AI label may be required when content has "photorealistic video or realistic-sounding audio" that was digitally created, modified or altered, including with AI. Meta's examples of content that needs a label:

• A video that appears realistic of a group of people walking around an outdoor market
• An audio file of two people talking
• A song created using AI-generated vocals
• A reel narrated with a realistic AI-generated voiceover

That last example covers most faceless Reels. Current AI narrators are built to sound human, so a story or explainer read by one falls under the rule even when every image is illustrated. What Meta doesn't require a label for: still images (they can still get one if Meta's systems detect AI) and video in a cartoon-like style, such as its example of an outdoor landscape drawn in a style resembling a cartoon.

Adding the label

On Instagram the switch sits in the creation flow. For a Reel or a feed post, turn on Add AI label before you share. For a Story, tap Options, then Label AI. Instagram's help page lists these steps for the iPhone and Android apps and says the feature isn't available on computers. You can also switch on AI-generated profile under Edit profile, which labels the account and its unlabelled posts, but Meta doesn't say it replaces the per-post label. Our Instagram Reels guide covers the rest of the posting side.

Facebook has the same requirement, but we couldn't find step-by-step instructions on an official Meta page. Look for the AI label option in the post or reel settings before you publish.

Auto-labels and penalties

Content made with Meta's AI tools, and content carrying industry-standard AI signals from other tools, may be labelled automatically ("AI info"). If the signal says the file was only edited with AI, the information sits in the post's menu instead of on the post itself. Meta's wording on enforcement is brief: "There may be penalties if you do not label content as required."

The EU AI Act: what changed on 2 August 2026

Article 50 of the EU's AI Act, the transparency article, has applied since 2 August 2026. Its paragraph 4 says deployers of an AI system that generates or manipulates image, audio or video content "constituting a deep fake" shall disclose that the content has been artificially generated or manipulated (Regulation (EU) 2024/1689). Three details decide whether that touches a faceless channel.

1. Are you a deployer? The Act doesn't cover people using AI for a purely personal, non-professional activity. The European Commission's Q&A on Article 50 says that when someone gains an economic benefit from the activity on a regular basis, it counts as professional and they're a deployer. A channel that earns money regularly fits that description better than the personal one.
2. Is it a deep fake? The Act's definition is AI content that resembles existing persons, objects, places, entities or events and "would falsely appear to a person to be authentic or truthful". An illustrated history video doesn't meet it. A photorealistic "archive photo" of a real event, or a realistic clone of a real person's voice, can.
3. Where is it seen? The Act covers deployers in the EU and, under Article 2, deployers outside the EU when the AI output is used in the EU.

The same Q&A adds that the disclosure must be visible or audible at first exposure, so invisible machine-readable marks don't count on their own. For a deep fake inside an evidently artistic, satirical or fictional work, it can be light enough not to spoil the work. Content generated before 2 August 2026 doesn't need labelling after the fact. National authorities enforce the rules, with fines of up to 15 million euros or 3% of worldwide turnover. If your business depends on EU audiences, get proper legal advice. Otherwise the takeaway is simple: a video realistic enough to need a platform label is also the one where a short on-screen "AI-generated" note costs you nothing.

Faceless formats, one by one

Here is how the rules land on the formats faceless channels actually make. "Yes" means that platform's rules call for a label.

AI-narrated history

The style decides it. Watercolor, anime or painted scenes fall under TikTok's "artistic styles" exemption and YouTube's non-realistic examples, and a stock AI narrator is the generic text-to-speech TikTok exempts. Photographic scenes of the past are different: YouTube names "a realistic scene that didn't actually occur", and a realistic fake "photo" of a historical moment is exactly the kind of image that gets screenshotted and shared as genuine. For history channels, illustrated styles are the cleaner choice; if you want the photographic look, label it.

Story time and explainers

A story read by a stock AI voice over gameplay or illustrations needs no label on TikTok or YouTube, which lists gameplay footage among its no-disclosure examples. The same goes for diagrams and infographics in a science or money explainer. Instagram and Facebook still want a label for the narration. The line moves on every platform once an image could pass as documentation: YouTube's must-disclose examples include a realistic tornado moving toward a real city that never happened.

Your own cloned voice

YouTube exempts "cloning one's own voice to create voice overs or dubs". TikTok has no own-voice exception and asks for disclosure when AI audio mimics the voice of a real person, so labelling is the safe reading there. Meta's realistic-voiceover rule applies whoever the voice belongs to.

Talking avatars and AI music

A realistic person speaking to camera who never recorded those words is what every one of these policies has in mind. In Blipix's Talking Avatar template, if you don't choose a face, the app generates a photorealistic presenter, so treat every avatar video as one that needs a label. The same applies to the AI music option in the Music template, which generates a full song with vocals: Meta lists songs with AI vocals, and YouTube asks for disclosure when AI music is the main focus. Instrumental background music under a narration isn't that case.

Voice clones and look-alikes of real people: the high-risk case

Cloning the voice of a celebrity, a politician, a well-known narrator or a historical figure with surviving recordings, or generating a realistic likeness of one, needs a label everywhere. TikTok requires disclosure when "AI-generated audio mimics the voice of a real person", YouTube when content "makes a real person appear to say or do something they didn't do", and Meta's realistic-audio rule covers it too. Under the EU AI Act, a realistic clone of a real person is the textbook deep fake.

Then the label stops being enough:

• TikTok doesn't allow the likeness of private figures without their permission, or public figures shown endorsing products, taking political stances or commenting on issues they haven't addressed, labelled or not.
• YouTube lets people request removal of realistic AI content that looks or sounds like them through its privacy complaint process. Whether the video was disclosed is one factor YouTube weighs, alongside realism, parody or satire, and whether a public figure is shown in a sensitive situation. YouTube also runs a likeness detection tool that helps enrolled creators find AI videos of their face, and it's working on voice.

Our rule of thumb: only clone a voice you own or have written permission to use, and label every video that uses it.

Posting from Blipix: which labels are set for you

We checked our own publishing code so you know exactly what happens when a video leaves Blipix. Posting from the editor or from a series needs a paid plan.

• TikTok: every video Blipix posts, from the editor or a series, goes out with TikTok's AI-generated content flag switched on, so TikTok shows its "Creator labeled as AI-generated" label. You don't need to do anything, and the label is accurate because everything Blipix makes is AI-generated. More on that in our TikTok video generator overview.
• YouTube: uploads are set as not made for kids and go out as Public, Unlisted (the Friends option) or Private, whichever you pick. Blipix doesn't answer YouTube's AI use question for you. For a video that needs disclosure, upload it as Private, open its details in YouTube Studio, set AI use to Yes, then publish it. In a series, that means choosing Private and publishing each realistic episode yourself.
• Instagram: Blipix posts Reels with your title and description as the caption, and it doesn't attach Meta's AI label. Since Meta counts a realistic AI voiceover as content that needs one, most narrated Reels do. For now, the reliable route is to download the video from the editor and post it from the Instagram app with Add AI label switched on.
• Facebook: Blipix doesn't post to Facebook, so you upload the file yourself and the labelling step happens there.

FAQ

Does an AI label hurt reach?

The platforms say no. TikTok's help page says the AI-generated content setting won't affect distribution as long as the video follows the Community Guidelines, and YouTube says disclosure alone won't limit a video's audience or its eligibility to earn money. Meta's pages don't address it either way. Monetization on YouTube is decided by separate rules on inauthentic, mass-produced content, not by the label. If a new account gets almost no views, the cause is usually elsewhere; our account warm-up guide covers the usual suspects.

Is a hashtag or a line in the caption enough?

On TikTok, yes: its guidelines accept a clear caption, sticker or watermark as well as its own label. YouTube asks for the AI use setting in Studio, and Meta describes its Add AI label toggle; neither says a caption replaces them. Adding "AI-generated" to the caption on top of the official setting never hurts.

Can I add a label to a video I already posted?

On YouTube, yes: the AI use answer sits under Attributes in the video's details in Studio. TikTok's and Instagram's help pages only describe labelling at the moment you post, so get it right before you publish there. Under the EU AI Act, content generated before 2 August 2026 doesn't need labelling after the fact.