Quick Answer
For a marketing campaign with AI video, the ethics come down to five steps. Decide whose face and voice appear, and get written consent that names AI cloning and the ad uses for anyone who is not you. Never make a real person, customer or public figure appear to endorse something they did not. Check the licence on the plan you will pay for. Label realistic output where you publish: TikTok, YouTube and Meta each require it, and the EU AI Act has required deepfake content to be disclosed since 2 August 2026. And keep the consent and source files with the campaign. This is information, not legal advice.
Five steps for an AI video campaign, from whose likeness appears to labelling on TikTok, YouTube and Meta, with the rule behind each and the EU AI Act duty in force since 2 August 2026. Checked 16 September 2026.
Related next steps
Marketers get asked which AI video tools are ethical. The better question is which steps in the campaign carry the risk, because they are the same whichever tool renders the video. Here they are in the order a campaign meets them, with the rule behind each, read on 16 September 2026. It is information rather than legal advice.
The five steps
- Decide whose likeness appears. Your own team, a hired presenter, a customer, or a generated character that is nobody. Each choice sets what follows.
- Get consent in writing for anyone who is not you. Name AI cloning, the channels, the duration and what happens when the person leaves or withdraws. Tennessee's ELVIS Act, in force since 1 July 2024, treats a simulation of a voice as the voice and prohibits use without authorization.
- Never manufacture an endorsement. Making a real person appear to recommend your product is exactly the conduct the rules target: the FTC proposed a rule in February 2024 ↗ prohibiting impersonation of individuals, and the federal NO FAKES Act of 2026 would create a right against unauthorized digital replicas, though it was only reported to the Senate on 24 June 2026 and is not law.
- Check the licence on the plan you pay for. Paid is not the same as commercial: D-ID's Lite plan is paid and personal use only. Percify carries commercial rights on every plan.
- Label where you publish, and keep the records. See below.
Labelling, by platform
- TikTok requires a label ↗ on realistic AI content, including AI generated speech, and an automatic label cannot be removed.
- YouTube requires disclosure ↗ through the 'AI use' setting when AI makes a real person appear to say something they did not.
- Meta has required disclosure ↗ of photorealistic video or realistic sounding audio since 6 February 2024, shows an "AI info" label, and says it may apply penalties when people do not disclose.
- The EU: Article 50 of the AI Act ↗ requires deepfake content to be disclosed, and has applied since 2 August 2026.
Keep the signed consent, the source photo and voice sample, and the script together with the campaign, so any question about the ad can be answered with documents rather than memory. The per platform table is in each platform's label rule.
Choosing the tool
The tool matters less than these steps, but its published terms still matter: how to judge an AI video tool from its terms goes through licence, marking and consent. For ads specifically, TikTok ads with AI avatars covers the formats; for the legal detail, voice cloning and the law; for the general duties, ethical AI video.
Ready to Create Your Own AI Avatar?
Make a talking video of yourself from one photo, in your own voice. No watermark on any plan.
Get Started FreeFree to start · no credit card
Got questions?
Frequently asked
By handling five steps: decide whose likeness appears, get written consent for anyone who is not you, never manufacture an endorsement from a real person, check the licence on the plan you pay for, and label realistic output where you publish while keeping the consent and source files. This is information, not legal advice.
Realistic ones do on the major platforms: TikTok requires a label on realistic AI content including AI speech, YouTube requires disclosure through its AI use setting, and Meta has required disclosure of photorealistic video or realistic audio since 6 February 2024. In the EU, Article 50 of the AI Act has required deepfake disclosure since 2 August 2026.
Only with their written consent naming AI cloning, the channels and the duration. Making a real person appear to endorse a product without authorization is the conduct the ELVIS Act prohibits and the FTC's proposed impersonation rule targets.
No. The NO FAKES Act of 2026 (S.4591) would create a federal right against unauthorized digital replicas of voice and likeness, but it was reported to the Senate on 24 June 2026 and had not been enacted as of 16 September 2026.
No. D-ID's Lite plan is paid and licensed for personal use only. Check the licence on the exact plan; Percify carries commercial rights on every plan, including free.
