Yes, when your video shows something realistic that AI made or changed, such as a person, a place, an event or a voice. Instagram and Facebook, TikTok and YouTube all require their AI label in that case, while scripts, captions, colour grading and cartoon-style animation need no label anywhere.
Yes. If a Reel, TikTok or Short shows something realistic that AI generated or changed, every one of these platforms expects you to turn on its AI label before you post. The test is realism, not the tool: an AI-written script or a cartoon needs no label anywhere, while AI footage of your storefront or an AI presenter does.
No Canadian law makes a business do this yet. Below are the platforms' own rules, read on their official pages on October 19, 2026, plus the advertising law that covers every ad. It is a summary, not legal advice.
What does "realistic" mean to each platform?
Instagram and Facebook. Meta's help page, Label AI content on Instagram, requires a label on photorealistic video or realistic-sounding audio that was digitally generated or altered, including with AI. AI images are exempt from the requirement, though Meta may label them itself.
TikTok. The Community Guidelines on edited media and AI (current version in force since September 24, 2026) require a label on AI-generated or significantly edited content showing realistic-looking scenes or people. Small edits, artistic styles like anime and generic text-to-speech are exempt.
YouTube, Shorts included. The page on disclosing GenAI content asks for disclosure when a real person appears to say or do something they did not, footage of a real event or place is altered, a realistic scene never happened, or AI music carries the video. Filters, captions, AI-written scripts and your own cloned voice are listed as fine without it.
Notice what none of them ask: whether you used AI at all. They ask whether a viewer could mistake something invented for something real. That is the question to put to every cut.
Which business videos need an AI label?
We checked common local-business situations against the three rulebooks. Where a platform does not address a case directly, the cell says so.
| In the video | Instagram and Facebook | TikTok | YouTube Shorts |
|---|---|---|---|
| Real footage with a colour grade, captions or an AI-written script | No (see note) | No | No |
| Cartoon or anime-style animation, even if AI drew it | No | No | No |
| AI b-roll of a real place: the skyline, your street, your storefront | Yes | Yes | Yes |
| A realistic AI presenter or "customer" who does not exist | Yes | Yes | Yes |
| A realistic AI voiceover in a generic voice | Yes | No, generic text-to-speech is exempt | Not on its list of required cases |
| An AI dub of the owner's own voice into another language | Yes | Yes, it mimics a real person's voice | No, your own cloned voice is exempt |
| A sign, car or person removed from real footage with AI | Yes, by its wording | Yes, if it misleads | Yes, if it meaningfully alters a real place |
| A photoreal 3D render of a building not yet built, no AI used | Yes, by its wording | Not covered unless AI made it | Not covered unless AI made it |
| An AI-made result or testimonial | Do not post it | Do not post it | Do not post it |
The note: read literally, Meta's "generated or altered" would catch routine edits, but its examples are all invented scenes and voices, and AI editing signals land in the post's menu, not on the post. Our reading: label when the change would alter what a viewer believes happened.
The render row surprises developers. Meta says the requirement applies when a digital creation tool or an AI tool made the realistic video, so a photoreal walkthrough of a suite that does not exist yet fits, even with no AI in the pipeline. TikTok and YouTube wrote their rules about AI, so a traditional render sits outside them, but a caption saying it is a rendering costs nothing. Renderings have a bigger problem anyway: buyers scroll past them, as our pre-construction video page explains.
Why does an AI voice need a label on Instagram but not on YouTube?
Each platform drew the line somewhere else. Meta's own examples of content that needs a label include a Reel narrated by a realistic AI voice. TikTok exempts generic text-to-speech that does not sound like a known person, but wants a label once AI audio imitates a real one. YouTube exempts your own cloned voice for voiceovers and dubs.
For a Toronto business posting in two languages, that split is practical. We shoot in English and Farsi, and an AI dub of an owner's own voice would be label-free on YouTube while TikTok and Instagram would both want it labelled. Since we deliver a separate cut for Instagram Reels, TikTok and YouTube Shorts, the label is decided per upload, not once per video.
Is animation the safe way to use AI visuals? What 240 viral videos show
Our outlier dataset holds 240 Reels and TikToks from September and October 2026 that beat their own account's median views. Nine of them, across five industries, are built on animation or 3D: four from dentists, two from med spas, and one each from real estate, a barbershop and a restaurant.
The format held its own. Those nine had a median multiplier of 35.9x, close to the 40x median of all 240, and four of the nine beat their account's median by 100x or more, against one in three across the whole set. The biggest, an animation of teeth grinding through the night, did 4.4 million views at 1,064x from an account with about 2,000 followers (October dentist report). A med spa video that showed microneedling in 3D before cutting to real footage did 6.9 million at 258x.
Two limits. Nine videos is a small sample, so treat this as a lead, not a law. And our notes describe what is on screen, not how it was made, so we cannot say whether any of these used AI tools.
Why it still matters: content that clearly reads as animation sits on the no-label side of every rulebook above. If you want AI to show what happens inside a tooth, a wall or an engine, a style nobody could mistake for footage skips the label question, and on this small evidence it travels about as well as anything else.
Does the AI label hurt reach?
TikTok and YouTube both say no. TikTok's Creator Academy page says the AI-generated content setting will not affect distribution as long as the video follows the guidelines. YouTube says disclosure alone will not limit a video's audience or its eligibility to earn money.
Meta does not say either way for the post label. Its page on AI-generated profiles says the account-wide label neither helps nor hurts reach, but accounts Meta believes are AI-generated profiles without it can lose eligibility for recommendations. Any brand running an AI mascot as a recurring character should read that twice.
One lever sits with viewers: in November 2025 TikTok said it would test a control that lets people see less AI-generated content in For You. Our dataset does not record labels, so we cannot say whether labelled videos do better or worse. Anyone quoting a number on that should show its source.
What happens if you skip a required label?
- Instagram and Facebook: Meta says there may be penalties, and it may apply the label itself when it detects AI.
- TikTok: unlabelled content may be removed, restricted or labelled by TikTok's team, depending on the harm. TikTok also says you cannot edit or remove the label after posting, so decide first.
- YouTube: creators who consistently fail to disclose can have a label applied for them, lose the video or be suspended from the YouTube Partner Program.
Skipping it may not hide anything. These platforms read Content Credentials, the C2PA metadata many AI tools attach to their files. TikTok began auto-labelling content that carries them in May 2024, and by November 2025 said its labelling had covered more than 1.3 billion videos. YouTube says a label applied because of C2PA metadata cannot be changed by the creator. If your AI tool writes those credentials, assume the label will show.
The same technology now works the other way: Instagram's Captured with a camera label and YouTube's matching disclosure verify real footage, though both allow little or no editing, so a fully edited Reel is unlikely to carry one.
Do ads follow different rules?
Yes, and since June the label on a Meta ad no longer depends only on you.
| Platform | What the ad rules add |
|---|---|
| Meta | A June 1, 2026 update says Meta detects ads made or edited with third-party AI tools and adds an AI info label under "About this ad", which may vary by region. Social issue, election and political ads must also disclose photorealistic AI or be rejected. |
| TikTok | The ads policy on misleading content (updated April 2026) allows AI-generated or significantly edited ads only with the AIGC label or your own clear disclaimer. Undisclosed AI gets the ad rejected or restricted. Lighting and background changes count as minor. |
| YouTube | Google's political content policy requires a synthetic content checkbox on election ads and adds the disclosure itself on Shorts placements. |
Is labelling AI the law in Canada?
Not for businesses, not yet. The federal government's AI transparency consultation paper (modified August 13, 2026) states that Canada does not currently require developers or deployers of generative AI to disclose AI-generated content. The consultation closed September 23, 2026.
The bill to watch is C-34, the Safe Social Media Act, introduced June 10, 2026 and still at second reading when we checked on October 19. As proposed, it would require regulated social media operators to label synthetic content likely to be mistaken for a real recording. The duty would sit with the platform, not the business posting.
What applies today is advertising law. The Competition Act bans promoting a product or business with materially false or misleading claims, in any medium. The Competition Bureau's section of a 2025 paper by Canada's digital regulators uses deepfaked celebrity endorsements and fake reviews as examples that could raise issues. The paper says it is not a policy position, and no court has ruled on those examples.
Ad Standards, the industry's self-regulator, went further in the Fall 2025 update to its influencer disclosure guidelines: say an AI influencer is virtual, never let it give a testimonial about something it cannot experience, and disclose product images AI generated or significantly altered. Its sharpest line: "A disclaimer cannot correct a misleading impression."
If you also market in the US, the FTC rule on reviews and testimonials, announced August 14, 2024, bans fake ones, naming AI-generated reviews from people who do not exist.
So today the label is a platform rule. The legal risk is an AI video that misleads: a result that never happened, a customer who does not exist, a face used without permission.
A checklist before you post
- Could a viewer mistake anything here for something that happened? Label it.
- Check the voice on its own; the platforms disagree.
- Switch the label on at upload: Add AI label (Instagram), More options (TikTok), Attributes then AI use (YouTube Studio).
- Add "AI-generated visuals" to the caption too.
- Never generate the result: the before-and-after, the finished kitchen, the new smile.
- No AI customers, reviews or testimonials.
- Get consent for every real face and voice.
- Running it as an ad? Disclose in Ads Manager too.
- Keep the project file in case real footage gets labelled by mistake.
Captions for people watching without sound are a separate legal question in Ontario, covered in our AODA captions guide.
Frequently asked questions
Do you have to label AI content on Instagram?
Yes, if a Reel or video contains photorealistic video or realistic-sounding audio that was digitally generated or altered, including with AI. AI images do not require it, though Meta may label them itself. Turn on the Add AI label toggle when you create the post or Reel.
Why does my Instagram post say AI info?
Meta renamed its "Made with AI" label to "AI info" in July 2024. It shows on the post when you disclose AI or Meta detects AI-generated content. When an AI tool only edited the file, the information sits in the post's menu instead.
How do you label AI on TikTok?
Before posting, tap More options and turn on the AI-generated content setting, or add your own clear caption, sticker or watermark. TikTok says the label cannot be edited or removed after posting and does not affect distribution when the video follows the guidelines.
Does YouTube require AI disclosure on Shorts?
Yes, the same rule covers every upload. In YouTube Studio, under Attributes, set AI use to Yes when the video shows realistic AI content such as a real person saying something they did not, altered footage of a real place, or a scene that never happened. YouTube says disclosure alone does not limit reach or monetisation.
Is it illegal not to label AI in Canada?
No Canadian law requires a business to label AI in a social video as of October 19, 2026, and Bill C-34 would put a labelling duty on platforms, not posters. The Competition Act already bans materially misleading advertising, so an AI video that misleads customers is a legal risk with or without a label.