Straight answer: there is no single official name. Depending on who’s talking, you’ll hear “AI video,” “synthetic media,” “generative video,” “deepfake,” or “digital human content,” and the term someone picks usually tells you whether they’re trying to sell it to you, disclose it to you, or warn you about it.
I found this out the hard way about eighteen months ago, when a client’s legal team stopped a project because of one word in a brief. Not because the video was bad. Because I’d written “AI presenter video” and their compliance officer read that as marketing spin for what she called, flatly, “a deepfake of our CEO.” Same footage. Wildly different reaction depending on the label. That’s the whole problem with this topic in one sentence.
Why there’s no single agreed term
AI-generated video is a young category built by three different groups who don’t talk to each other much: engineers, marketers, and regulators. Each group named it for their own purposes.
- Engineers and researchers call it generative video or text-to-video, because that’s literally what the model does, it generates pixels from a prompt.
- Marketers and tool companies call it AI video, AI avatar video, or digital human video, because those terms sound like a product feature rather than a technical process.
- Regulators, journalists, and platforms call it synthetic media or, when it involves a real person’s face or voice without full consent, a deepfake. This is the label with the worst reputation, and for good reason. It started as the name for non-consensual face-swapped content and it never fully shook that off.
So when a client asks me “what do we call this,” my honest answer is: it depends who’s asking and what you’re trying to signal to them.
The specific terms you’ll run into
AI video / AI-generated video
The catch-all. Used by tools like Synthesia, HeyGen, and Runway, and by most marketing writers, including me most of the time. It’s vague on purpose. It covers everything from a fully synthetic talking-head avatar to a five-second AI-extended background in an otherwise real filmed video.
Synthetic media
The term you’ll see in policy documents, academic papers, and platform trust and safety pages. Meta, YouTube, and TikTok all use some version of this in their disclosure policies rather than “AI video,” because “synthetic” is broader and covers audio-only and image-only manipulation too, not just video.
Digital human / AI avatar
This is the term the corporate training and internal communications world has settled on. If you’ve seen a company use a repeatable AI presenter for onboarding videos or product explainers, that’s usually what it’s being called internally: a digital human, not a deepfake, even though technically the underlying tech is close cousins with the same generative video models.
Deepfake
Technically correct for a lot of face-and-voice AI video, but socially radioactive. Nobody selling a legitimate product calls their own output a deepfake. It’s the term used almost exclusively by critics, journalists covering fraud, and regulators writing law. If a client uses this word about your work, take it seriously, it means they’re worried about consent, not quality.
Text-to-video and generative video
The technical, model-focused terms. Used in research papers, model release notes (Sora, Veo, Runway Gen-4, Kling), and by people who care about how the video was made rather than what it’s for. If you’re comparing tools for your own content creation stack, this is the terminology you’ll see in the technical comparisons and benchmark write-ups, not the marketing pages.
The uncomfortable part nobody in the “AI video terminology” articles wants to say plainly
Here’s the bit that gets glossed over: the softer terms exist partly to make the content sound less risky than it legally is. Calling something a “digital human” instead of a “synthetic likeness” doesn’t change your disclosure obligations. It just changes how nervous your legal team feels reading the brief. I’ve watched agencies pick the friendliest possible name for exactly this reason, and it works right up until a client’s compliance department, a platform’s trust and safety team, or a new law forces the plainer word back onto the table.
That’s happening now. The EU AI Act’s transparency rules for AI-generated and manipulated content apply from 2 August 2026, and they don’t care what your marketing deck calls it. If a video shows a real or realistic-seeming person doing or saying something they didn’t do or say, it has to be labelled as artificially generated or manipulated, full stop. The friendly branding doesn’t buy you an exemption.
What platforms call it in their labels
This is the part with the most practical use, because it’s where you’ll get flagged or not flagged.
- YouTube: requires creators to disclose “altered or synthetic content” that’s realistic, and shows a label reading “Altered or synthetic content” on the video itself for sensitive categories (health, news, elections, finance).
- Meta (Facebook and Instagram): uses “AI info” as the visible label and refers to the underlying detection as identifying “AI-generated” content, sometimes flagged automatically from C2PA metadata baked into the file by the generating tool.
- TikTok: requires creators to tag realistic AI-generated content with the “AI-generated content” label in the posting flow, and calls the broader category “synthetic media” in its community guidelines.
If you’re building anything for social first, especially on Instagram, this labelling matters more than the vocabulary debate. I’ve covered the mechanics of getting content seen and trusted on that platform in my Instagram help guide, and the short version is: unlabeled synthetic video that gets reported tends to get throttled hard, regardless of quality.
A real example: what I called it, and what I should have called it
Last year I ran a small test for a coaching client, three versions of the same product-explainer script: one filmed with a real presenter, one using an AI avatar tool, one using pure text-to-video with no human likeness at all. I labelled the brief “AI explainer variants” for internal use. Simple, functional, nobody blinked.
The moment it went external, the language had to change three times in one week:
- For the client’s marketing team: “AI-assisted explainer video.” Reassured them it wasn’t a wholesale replacement for their brand’s real presenters.
- For the client’s legal team: “synthetic video content, no real individual’s likeness used without consent.” That’s the exact phrase that got it approved, because it answered the actual worry, not the vibe.
- For the platform upload (LinkedIn): no formal label existed yet at the time, so we added a caption disclosure ourselves, “Created with AI video tools,” which is now roughly what most platforms expect as a minimum anyway.
Same eleven seconds of footage, three different names, because each audience was worried about a different thing. That’s the pattern to remember: the name isn’t really describing the technology, it’s answering the question the specific audience is silently asking.
How to pick the right term for your own use
Five steps, in order, and it takes about fifteen minutes once you know the checklist:
- Ask who’s reading the label. Internal team, client, legal, platform, or public viewer, each wants a different level of precision.
- Check if a real person’s likeness is involved. If yes, “synthetic media” or “AI likeness” is the honest term, not “AI video.” This is the line where deepfake territory legally starts.
- Check the platform’s own label wording. Match it. YouTube, Meta, and TikTok each phrase disclosure differently, and using their exact phrase avoids ambiguity with reviewers.
- Default to the plain, boring term in contracts. “AI-generated video” or “synthetic video content” holds up better than a brand-friendly euphemism if anyone ever has to read the contract back in a dispute.
- Keep a one-line disclosure ready for public posts. Something like “Made with AI video tools” covers you across most platform policies and most audiences’ basic expectation of honesty.
None of this needs to be dramatic. It just needs to be consistent, because inconsistency is what gets noticed and questioned.
Where this is heading
My honest read, after two years watching this space for clients: “AI video” will stay the everyday term because it’s short and everyone understands it. “Synthetic media” will win in policy and legal documents because it’s broader and more precise. “Digital human” will keep growing in corporate and training contexts because it sounds like a category, not a warning. And “deepfake” will keep meaning exactly what it started meaning, content made to deceive, whether the tool involved was a two-thousand-pound piece of software or a free app.
If you’re producing this kind of content for testimonials, explainer videos, or training, the label you choose in your contracts and briefs matters more than most people building their tool stack realise. It’s worth checking how established testimonial video production companies phrase their disclosures now, because most of the good ones have already updated their contracts for the EU AI Act deadline, and their wording is a decent shortcut to copying language that’s already been legally checked.
Frequently asked questions
Is “AI video” the same as “deepfake”?
Not quite. “AI video” is the broad, everyday term for any video made or significantly altered by AI tools. “Deepfake” specifically means a video that uses someone’s real face or voice, usually without full consent, and it carries a negative, deception-focused connotation that “AI video” doesn’t.
What do YouTube and TikTok officially call AI-generated video?
YouTube labels it “Altered or synthetic content” in its disclosure settings, while TikTok uses “AI-generated content” as its tagging option and refers to the wider category as “synthetic media” in its guidelines.
Do I legally have to disclose that a video is AI-generated?
In the EU, yes, from 2 August 2026 the AI Act requires disclosure for realistic AI-generated or manipulated video showing real or seemingly real people. Several platforms already require self-disclosure for realistic synthetic content regardless of location, so it’s safest to label it everywhere, not just where it’s mandatory.
What’s the difference between “synthetic media” and “AI-generated content”?
“Synthetic media” is the broader technical and policy term covering AI-altered audio, images, and video together. “AI-generated content” usually refers more narrowly to material created from scratch by a generative model, rather than real footage that’s been edited or altered with AI.
