AI-Generated Voice in Ads: When Is Disclosure Required?

The IAB AI Transparency and Disclosure Framework sets out exact guidelines dictating when AI-generated audio in advertising requires consumer disclosure — and when it does not. If your team buys or produces podcast spots, streaming audio placements, or voice-oriented AI creative, understanding these boundaries is critical for compliance.

Here is an in-depth look at precisely which AI audio scenarios trigger the disclosure requirement, the scenarios that do not, and the delivery standards you must follow according to the IAB's January 2026 framework.

What the IAB Framework Covers Under AI-Generated Audio

The IAB framework's rules apply to a broad range of AI audio use within commercial contexts and digital advertising. The framework explicitly encompasses several categories: synthetic speech, voice cloning, AI-generated music, sound design, and audio effects.

Because the definition is expansive, it means the rules are not limited to voiceovers alone. Any AI-generated or AI-modified audio element in an ad that falls into the trigger categories requires careful review against the guidelines.

For full context on how the audio rules sit within the broader IAB ecosystem — including image, video, and synthetic person requirements — jump to the IAB AI Disclosure Framework: Complete Advertiser Guide.

When AI Audio Disclosure Is Strictly Required

The framework requires explicit disclosure in two specific audio scenarios involving human representations.

1. Synthetic Voice of a Deceased Person

Disclosure is required when an ad uses AI-generated voice content that creates new speech or statements from a deceased individual — meaning words they never actually spoke while alive. Under the framework, this disclosure is required even if the individual's estate has granted explicit authorization.

There is one important boundary in the framework regarding the deceased: this rule does not apply to the use of archived, actual recordings where the person genuinely spoke those words during their lifetime. The trigger for disclosure is the creation of new AI-generated speech, not the integration of authentic archival recordings.

The required compliance label is exact: "AI-generated voice"

The framework's placement guidance for this content dictates that the verbal disclosure must appear before or after the AI-generated segment so consumers are immediately aware of the synthetic nature.

Example: A watch brand creates an ad using an AI-generated voice clone of a deceased celebrity endorsing new styles (with estate approval). Because new conversational segments were synthetically generated, disclosure is required.

2. Synthetic Voice of a Living Person Making Fabricated Statements

Disclosure is required when AI-generated voice content of a living individual makes statements that convey or imply real-world facts, experiences, personal actions, or documented events that the individual did not actually do, experience, or state. This applies even if the living individual gave authorization for the AI voice.

The framework draws a clear and important line here: this disclosure trigger is entirely distinct from standard scripted commercial endorsements. Promotional brand messaging explicitly authorized by the individual (e.g., standard voiceover scripts about product performance) does not trigger this rule. It is triggered when the AI voice is used to fabricate statements about historical or real-world specificities that never occurred.

The required compliance label remains exactly the same: "AI-generated voice"

The placement guidance requires a verbal disclosure before or after the content, or, in video formats, a text overlay that satisfies the visibility requirement.

Example: An automotive brand securely uses an authorized AI voice clone of a current racing champion stating: "I set the track record at Daytona using this exact model." If the driver never actually drove the vehicle or set the record, the scenario involves a fabricated real-world event and requires disclosure.

When Disclosure Is Not Required

The framework relies on a core threshold test to determine when AI audio is exempt from disclosure.

The Underlying Threshold Test

The framework states a threshold principle that applies across all audio decisions: if audio production or editing could be consistently performed with traditional tools — such as Logic Pro, Pro Tools, or mixing software — without requiring a consumer disclosure, then using an AI system to perform the identical task does not trigger the mandate. The threshold strictly scrutinizes whether the AI creates a false impression about who is speaking or what was said, not merely the mechanical process of how the audio was mastered.

1. Generic Synthetic Voices

AI-generated voiceovers that do not impersonate any specific living or deceased individuals and are used for standard advertising narration — where the human identity of the speaker is immaterial to the message — completely bypass the disclosure requirement.

Example: A SaaS software company leverages a generic text-to-speech AI narrator to walk through dashboard features in an instructional ad sequence. Since no identity is implicitly spoofed or falsified, no disclosure is required.

2. Authorized Voice Clones Reading Standard Scripts

When a living individual explicitly authorizes the use of AI to clone and replicate their voice for standard advertising purposes (e.g., product endorsements, brand messaging, basic scripted voice acting), disclosure is not necessarily triggered. In the eyes of the framework, this is treated equivalently to a standard, authorized human voiceover performance. Disclosure is only triggered when that clone claims false past events as true.

3. AI Soundscapes and Background Audio

AI-generated music, ambient sound, or sound effects where no reasonable average consumer would incorrectly interpret the content as being performed, composed, or created by a specific artist do not require any specialized disclosure labels.

Example: A travel application dynamically generates calming background wave and wind effects using prompt-guided audio tools behind a flight deal advertisement. Since consumers do not expect this to be the creative work of an identified artist, no disclosure is needed.

4. Technical Mastering and Engineering Practices

Tools utilizing AI logic to execute noise reduction, track equalization, volume normalization, and intelligent mixing — technical enhancements that do not alter the substance of what was recorded — are fully exempt.

Additionally, AI applied to execute minor voice corrections such as automated breath removal, pacing and timing adjustments, intelligent de-clicking, and basic pitch correction fall within universally accepted audio engineering practices and do not require disclosure.

How to Deliver AI Audio Disclosure Formatting

If your audio fits the requirements for disclosure under the IAB framework, how you place the notification depends on the advertisement's format.

Audio-Only Placements (The 60-Second Rule)

For audio-only environments lacking a visual interface — such as linear radio spots, native podcast mid-rolls, streaming audio ads, and smart speaker insertions — the framework mandates strict auditory cues.

Disclosures must be clearly spoken aloud in the exact same language as the primary advertisement. They must be spoken at a normal, intelligible pace. The framework dictates that the disclosure phrase ("AI-generated voice") must be positioned so that it naturally occurs before or immediately following the synthetic segment.

If the audio advertisement exceeds 60 seconds in complete length, the disclosure is subject to a repetition mandate: it must be spoken originally and then repeated at least once during the ad flight.

• Ad 60 seconds or under: Disclose once.
• Ad over 60 seconds: Disclose at least twice (the original disclosure plus at least one repeat).

Audio Within Video Placements

Where video is present directly alongside AI-generated audio (e.g., a Connected TV spot, pre-roll, or social video), the delivery standards are bifurcated:

• For living persons making fabricated statements: The framework allows brands to rely on a clearly legible on-screen text overlay as the primary disclosure method, replacing the need for a verbal cue.
• For the voice of a deceased person yielding new speech: A verbal disclosure remains the strictly specified method regardless of visual availability.

Frequently Asked Questions

Related Reading