Yes. A cloned or synthetic voice that sounds like a real person is audio content that "would falsely appear authentic", the core of the Article 50(4) deepfake duty. That covers a clone of a specific person's voice and also a fully synthetic voice realistic enough to be taken for a human speaker. Deceptive intent is not required; a friendly AI voiceover is in scope simply by sounding real.
The disclosure format for audio, per the June 2026 Code of Practice: a spoken disclaimer at the start of the audio, repeated if the piece runs longer than 30 seconds. A note in the ad's description text or your website footer doesn't reach the listener and doesn't do the job.
For AI voices in interactive settings (outbound calls, voice bots), you're additionally in Article 50(1) territory: the system is interacting directly with a person, so the call must open with a statement that the caller is dealing with an AI assistant, before any substantive content. The two duties stack; one opening line can be written to satisfy both.
Cloning a real, identifiable person's voice raises a second, independent body of law the AI Act doesn't displace: personality and image rights, GDPR (a voice is personal data), and, for actors and presenters, contract. Consent from the person cloned is a floor, and it does not substitute for the AI Act's disclosure to listeners; nor does the disclosure substitute for consent.
The artistic carve-out (attenuated, non-intrusive disclosure for clearly artistic or satirical works) does not cover primarily commercial content, so ads can't rely on it. For evidence, keep the audio master with its opening disclaimer and a dated note of where it ran.
Not sure how this applies to your specific setup? The free 2-minute check tells you which of the four Article 50 duty areas likely apply to your company. No email required.
Take the free check Generate disclosure text