Synthetic Voice and Likeness in Creator Ads
Review consent, usage scope, deception, testimonial, and synthetic-media disclosures before using a cloned voice or digital replica in an ad.
Before using a synthetic creator voice or likeness, verify consent and licensed use, then separately review whether the ad could deceive viewers and whether a synthetic-media disclosure is required. Disclosure does not cure unauthorized use. Permission to train or generate a replica also does not automatically authorize every script, platform, territory, or paid placement.
Define the replica and the authorized person
The U.S. Copyright Office's Copyright and Artificial Intelligence initiative uses “digital replica” for digital content that realistically but falsely depicts an individual. Its report explains that replicas can be authorized or unauthorized and can involve image, video, or audio.
Create a provenance record:
- Real person whose voice or likeness is represented
- Capture, training, or source material used
- Vendor and model that generated the output
- Person approving the final script and performance
- Whether the replica is obvious fiction or likely to appear authentic
- Every party that can generate, edit, store, or distribute it
Do not call a voice “generic” merely because a vendor generated it. Check whether it is readily identifiable as a particular creator, actor, employee, or celebrity.
Verify consent and contract scope before disclosure
Name, voice, likeness, publicity, labor, contract, privacy, and unfair-competition rights vary by jurisdiction. The license should answer practical questions:
- Which voice, likeness, or performance can be replicated?
- Which products and claims may the replica deliver?
- Which media, countries, languages, and paid placements are allowed?
- Can the brand edit, translate, or generate new performances?
- Who approves scripts and final renders?
- How long can the model and outputs be used and retained?
- What happens after termination, breach, death, or vendor change?
An ordinary creator-content license may cover recorded footage but not a model that generates unlimited new performances. Obtain qualified counsel; do not infer that uploading source audio transfers these rights.
Review deception and testimonial meaning
The FTC's Consumer Reviews and Testimonials Rule Q&A explains that AI avatars can create testimonial issues and that false use of a celebrity avatar can violate the Rule where consumers would think the celebrity actually gave the testimonial. The broader FTC Act can reach deceptive uses beyond the Rule.
Ask whether viewers could believe:
- The real creator spoke these exact words
- The creator personally used the product
- The creator holds a stated credential or opinion
- The creator approved this language and placement
- The depicted result actually occurred
Avoid first-person experience that the real person did not have. A small AI label does not neutralize a fabricated product result or unauthorized endorsement.
Use the sponsored-video review checklist for claims and media. The AI influencer disclosure guide covers fictional avatars that may not replicate an identifiable person.
Determine whether synthetic-media transparency applies
Article 50 of the EU AI Act includes duties for certain generated or manipulated content, including deepfakes, with definitions and exceptions that require specific analysis. U.S. states and platforms can impose other rules for digital replicas, election content, commercial likeness use, or AI labels.
Do not make one global “AI-generated” sticker the entire compliance plan. Map the audience, media, subject, product, and applicable rules. Disclosure presentation may need to be readable, audible, or machine-readable depending on the requirement.
Even when no specific AI label is required, a disclosure may be necessary to avoid the false impression that a real creator performed or endorsed the script. Test likely audience understanding.
Use a preflight checklist for every generated version
- Signed consent and license cover the exact use
- Script is within approved products and claim categories
- Real creator approved the final performance where required
- No invented experience, credential, or result
- Sponsor and commercial relationship are clear
- AI or digital-replica label is applied where required
- Captions and translations match approved meaning
- Vendor access and model retention follow contract
- Final output has a version ID and expiration date
- Takedown and revocation paths have been tested
High-risk uses deserve extra review: health claims, financial advice, minors, deceased personalities, sensitive contexts, and replicas designed to be indistinguishable from authentic recordings.
Preserve evidence and stop unauthorized generation
Keep consent, source provenance, vendor terms, model permissions, scripts, approvals, disclosures, final files, placements, and deletion evidence. Revoke credentials and vendor access at the end of the authorized period. Monitor for unapproved variations and respond under the contract and applicable law.
Laws in this area change rapidly. Refresh official sources and obtain jurisdiction-specific advice immediately before publication and reuse.
Tie every synthetic performance to an approved script
CherryBowl maps claims and disclosures in each final video to timestamps so reviewers can verify the exact authorized version.
See the AI review a videoOr join the early-access waitlist.
or book a call
General information only, not legal advice. Consent, publicity, AI transparency, contract, privacy, and advertising rules require jurisdiction- and fact-specific counsel.