Labelling AI in ads: a practical guide
Label AI when a viewer could mistake it for real. On a realistic AI ad, the IAB framework puts the label on the first frame and keeps it visible. Repeat it in captions and descriptions. It tells the audience the brand has nothing to hide.
Made with AIEvery film we make carries one line at the end: Made with AI. Directed by people. This guide to labelling AI in ads covers why we do it, what the industry framework actually asks, and where a label should go on each platform. Part of it corrects a habit of our own: a closing line alone is not what the framework asks for on a realistic AI ad.
This is practical guidance, not legal advice. Rules differ by market and they are moving.
Why labelling AI in ads protects the brand
Audiences notice AI. Finding it without being told can feel like being tricked. Being told plainly lets people judge the work on its quality.
A label also protects the people who sign the work off. If a film is questioned after launch, a marketing director wants to point to a clear decision made before launch, not explain why nobody decided.
What the IAB AI disclosure framework asks
The IAB published its AI transparency and disclosure framework in January 2026 and a second version in August 2026. It is voluntary. It does not ask you to label everything. Its test is materiality: could this use of AI mislead a reasonable consumer about what is authentic, factual or human-created?
In summary, from the August 2026 version:
| Disclosure called for | Disclosure not called for |
|---|---|
| Realistic video generated from prompts, even after human editing | Clearly stylised or fantastical content |
| AI avatars that could be mistaken for a real person | Routine post-production such as colour grading |
| AI replicas of deceased people, in any capacity | Generic AI figures in crowds and backgrounds |
| AI replicas of living people in events that never happened | Authorised AI replicas of living people in standard endorsements |
It also sets out how. Use a sparkle icon or a clear text label such as "AI-generated". For realistic synthetic video, the label appears on the first frame and stays visible. It should be readable, in the language of the audience, and included in captions. Behind the label, the framework adds machine-readable C2PA metadata, which travels with the file.
Where the AI label on video goes
We recommend four layers for a realistic AI ad. Each one covers a gap in the others.
- On screen from the first frame. Small, clean, high contrast, clear of platform buttons. This is the one the framework asks for.
- A closing line. Ours is Made with AI. Directed by people. It says who is accountable. It does not replace the first-frame label on a realistic ad.
- Captions and descriptions. Repeat the words in the post copy and the subtitle file.
- Metadata. Keep C2PA data in the delivered file where your tools and the platform support it.
Set the words in Arabic for an Arabic audience. Agree all of this in the brief, because a label added at the last minute tends to land on top of a super.
Made with AIAI ad disclosure, platform by platform
Platform settings change often, so treat this as a checklist and confirm each one at upload. The framework itself notes that platform practices vary and continue to evolve.
| Where it runs | What to do |
|---|---|
| YouTube | Answer the AI use question in Studio at upload. Keep the on-screen label and add the line to the description |
| Instagram and Facebook | Look for the AI marking in Ads Manager and at posting. Meta may add its own "AI info" label. Keep yours regardless |
| TikTok | Use the platform's AI-generated content setting. Keep the label above the bottom caption area |
| Snapchat | Burn the label into the file. Keep it inside the safe area for vertical video |
| Add the line to the post text. Keep the on-screen label | |
| Burn it in. Captions and descriptions do not reliably survive a forward | |
| TV and cinema | Burn it in, at a size that reads at viewing distance. Check the broadcaster's own rules |
| Your website | Label beside or on the player, and repeat it in the alt text and page copy |
The rule underneath the table is simple. A platform's label can be removed, cropped or missed. The one in the file stays with the film.
Public and historic figures
This is where labelling alone is not enough. Some AI films dramatise public or historic figures. Under the framework, an AI replica of a deceased person calls for disclosure in any capacity, even with estate authorisation. So does a replica of a living person placed in an event that never happened.
A label tells the audience what they are watching. It does not settle whether you may use the likeness. Take four steps before the script is approved.
- Get legal review for each market. Likeness, defamation and advertising rules differ. Ask counsel, not your studio.
- Check who else has a say. Families, estates, institutions and regulators may all matter.
- Be careful with words. Never put an endorsement, a claim or a new statement in the mouth of a real person without clearing it.
- Label from the first frame, and keep a record of what was approved and by whom.
If the answer is unclear, change the shot. Most films lose nothing by using a fictional character.
Does a label hurt the work?
It can, and it is better to say so. The IAB framework cites an NYU Stern study which found that telling consumers an ad was made with generative AI cut its click-through rate by 31.5%. That is the framework's own reason for reserving labels for real risk rather than labelling everything.
We still label every film, including ones the framework would let pass. That is a policy choice. We think a clear label costs less than being found out, and it is simpler than deciding film by film. We have no figures of our own to prove that, so test it on your own campaigns. Whatever you decide on stylised work, keep the label on anything a viewer could take for real.
A one-line policy you can adopt
"We use AI to make our films. We label every one. AI people are never presented as real testimonials, and a person approves every frame."
Each part is checkable, which is the point. Read how we apply it in our AI disclosure policy, or see how labelling fits the wider process in AI video production.
Labelling AI in ads questions
Does the law require an AI label on ads in Saudi Arabia or Egypt?
Ask counsel for your market and category, because rules are moving and this article is not legal advice. The IAB framework is voluntary and does not replace legal requirements.
Is a closing line enough to meet the IAB framework?
Not for realistic AI video. The framework's placement for that content is the first frame, staying visible. A closing line is a good addition to it.
Does a stylised or animated AI film need a label?
The framework does not call for one when no reasonable viewer would take it for real. Our own policy labels it anyway.
Can an AI person give a testimonial in an ad?
We do not present AI people as real customers, employees or witnesses. If someone is speaking from real experience, we recommend filming them.
