Does My AI Ad Need a Disclosure Label?

The Question Every Creative Team Is Asking Right Now

AI is now embedded in nearly every stage of ad production — generating lifestyle imagery in Midjourney, writing headline variants in ChatGPT, cloning a voice in ElevenLabs. Most creative teams using these tools have no consistent policy on when, or whether, to put a label on the finished ad.

That ambiguity just got a lot more expensive. The New York Synthetic Performer Law — carrying civil penalties of $1,000–$5,000 per violation — takes effect in June 2026. The EU AI Act's labelling provisions are phasing in. And platforms including Meta, TikTok, and YouTube already auto-apply AI labels under their own detection logic, independent of advertiser intent.

The IAB's new framework gives the industry a baseline to work from. It won't resolve every edge case — but it does establish a decision logic you can operationalize.

The Core Principle: Materiality, Not Volume

The most important thing to understand about the IAB framework is what it doesn't say. It does not say: "If you used an AI tool in production, you must disclose." That standard would be unworkable — AI is now in Photoshop, in Premiere, in every spell checker and caption generator on the market.

In other words: the question is not "did we use AI?" — it's "could a reasonable person watching this be misled about the nature of what they're seeing or hearing?"

The framework applies this through three evaluative criteria:

All three criteria point at the same target: the consumer's ability to make an informed assessment. Disclosure is required when omitting it would deceive a reasonable consumer about something material.

When AI Ad Disclosure Is Required: Rules by Content Type

The IAB framework applies the materiality test differently across five content types. Each has specific decision branches, specific exemptions, and specific label texts.

AI-Generated Images

The key variable is photorealism — not the tool used, not whether the image started from a real photograph.

A photorealistic AI render of a coffee cup on a kitchen counter requires disclosure. A clearly illustrated version does not. The trickiest edge case is substantive editing of real photography: AI changes that alter the appearance of the product itself or of a real person in the ad require disclosure. Background-only replacements are generally exempt.

AI-Generated Video

Video rules follow image logic but add a persistence requirement: when a label is required, it must remain visible for the entire video — not just the opening frame.

AI upscaling, frame interpolation, and noise reduction are all exempt — these are established post-production techniques a consumer neither expects to see disclosed nor would find material.

AI-Generated Audio and Voice

Audio rules are built around identity and authenticity, and they carry the sharpest legal exposure of any content type for U.S. advertisers. The framework's audio provisions map directly onto the New York Synthetic Performer Law.

The key distinction: an authorized AI voice clone of a living athlete reading a scripted endorsement is treated like a standard voiceover performance — no disclosure required. The same voice clone being made to describe a race they never competed in requires disclosure. The risk isn't the synthetic voice; it's the fabricated statement.

Synthetic Influencers and AI Personas

This is the highest-risk category for regulatory fines across both NY and California, and the one where brands are most likely to have compliance gaps they don't know about.

The deceased performer rule is a hard line with no exceptions — estate authorization, payment, or respectful intent does not remove the disclosure requirement. For AI chatbots, disclosure must appear at the moment the consumer begins interacting, not in a post-interaction screen.

AI-Generated Text and Copy

Unambiguous: the IAB framework does not require consumer-facing disclosure for AI-generated text. Headlines, slogans, product descriptions, email subject lines, social captions — none of these trigger a disclosure requirement. The principle that governs text is truth in claims, not authorship transparency: advertisers remain responsible for the accuracy of every statement, AI-written or not.

Quick Reference Table

What the Label Actually Has to Look Like

Getting the disclosure requirement right is only half the problem. The label itself has to meet the IAB's format standards — or it doesn't count.

The primary method is a text label using the specific IAB strings above (not paraphrases like "powered by technology"). The label must appear near the content, remain visible for the consumer's full exposure, use plain language, and provide sufficient color contrast to be legible.

When text labels would materially compromise the creative — constrained banner units, small format ads — the framework allows alternatives: platform visual indicators (the sparkle icon ✨, the C2PA Content Credentials icon, or platform-applied badges), hover/tap information icons in digital environments, or adjacent placement next to rather than on the creative.

Is the IAB Framework Actually Binding?

The framework itself is voluntary industry guidance — no regulatory body enforces it directly. But the framework was designed to align with the laws that are binding:

The New York Synthetic Performer Law (effective June 2026) requires conspicuous disclosure of AI-generated performers and carries civil penalties of Fine: Up to $5,000 per violation per occurrence. California SB 942 (August 2026) covers platforms with 1M+ users and requires embedded metadata disclosures. The EU AI Act Article 50 requires AI labelling as a consumer protection baseline, with full force in 2027. South Korea's Telecommunications Act already mandates blanket labelling of all AI-generated photos and videos in advertisements — no materiality test, no exemptions.

Compliance with the IAB framework doesn't guarantee compliance with all of these. But it gets agencies to a documented, defensible standard for most US campaigns — and provides the foundation for layering in jurisdiction-specific requirements. If your workflow involves C2PA metadata, the C2PA IAB Compliance Auditor checks your asset files for missing and inconsistent fields.

The Implied-Truth Effect: Why Getting This Wrong Cuts Both Ways

There's a counterintuitive risk worth flagging. Advertisers who apply disclosure labels responsibly can inadvertently create a disadvantage relative to those who don't: consumers tend to perceive unlabeled content as more credible simply because other content carries a disclosure. The IAB framework calls this the implied-truth effect.

This is precisely why the industry needs a consistent standard rather than voluntary opt-in. Inconsistent disclosure doesn't protect consumers — it rewards non-disclosure. The framework's goal is to establish a baseline where labeled content earns trust, and unlabeled content that should have been labeled faces genuine legal and reputational risk.

Frequently Asked Questions

Related Reading