What the Ad Industry’s AI Disclosure Problem Means for Affiliates

There is a working paradox sitting at the centre of AI advertising right now, and the industry has not agreed on what to do with it.

Research from NYU Stern and Emory University, published in late 2025, found that fully AI-generated ads outperform human-made ads by up to 19% in click-through rates. The same study found that telling consumers the ad was AI-generated reduced its effectiveness by up to 31.5%. The image did not change. The product did not change. The label changed, and one in three clicks disappeared.

That is the paradox. The machine makes better ads. Saying so makes them worse.

The study and what it actually found

The research, led by Hyesoo Lee at NYU and co-authored by Vilma Todri and Panagiotis Adamopoulos at Emory University alongside NYU's Anindya Ghose, tested three production approaches across a laboratory experiment and a live Google Ads field study: ads made entirely by humans, ads made by humans and then modified by generative AI, and ads generated from scratch by generative AI. The findings cut against how most marketing teams currently work.

Fully AI-generated ads outperformed both human-only and hybrid ads. The hybrid approach, where AI refines an existing human design, produced no meaningful improvement and sometimes performed worse. The researchers attributed this to what they call creative constraints: when AI modifies an existing design, it is working around decisions already made rather than generating from a clean brief. The result loses the visual processing fluency and emotional engagement that make fully AI-generated ads effective.

The disclosure finding was measured in the field. Adding a label stating that the ad was AI-generated or AI-edited cut click-through rates by approximately 1.17 percentage points compared to unlabelled human ads in absolute terms, which translates to the 31.5% relative reduction that has since been cited across the industry. The researchers did not prescribe a policy response, but the implication is clear: consumer scepticism toward AI-produced content is real and measurable in performance data, not just in survey responses.

Why this matters more for affiliates than for brand advertisers

A brand running display advertising at scale can absorb a 31.5% CTR reduction as a cost of compliance, offset it with media budget, and move on. For affiliates, the economics are tighter. Commission income depends on clicks converting. A material drop in click-through rate on AI-generated creative, caused by nothing more than a disclosure label, hits margin directly.

But there is a longer view that affiliates should not ignore.

Affiliate marketing runs on audience trust as its primary asset. Click-through rate measures immediate response. It does not measure whether a reader returns, whether they trust your next recommendation, or whether they associate your content with accuracy. A 31.5% drop in CTR on a single campaign is recoverable. A reputation for opacity about how content is produced is harder to walk back, particularly as audiences become better at identifying AI-generated visuals and increasingly hostile toward brands that obscure it.

The Wirecutter editorial transparency move in March 2026 illustrated this directly. As we covered at the time, a publisher that documents its methodology and is open about its commercial model gives readers, and AI search systems, more to work with when evaluating credibility. Disclosure does not have to be a liability. How it is framed determines whether it reads as a warning label or a mark of professionalism.

What regulators have actually said

The regulatory picture is getting clearer, though not uniformly so.

The EU AI Act's Article 50 transparency obligations require deployers of AI systems to disclose when content constitutes a deepfake, meaning AI-generated or manipulated image, audio, or video that imitates real people, objects, or events in a way that could mislead someone. For AI-generated text published to inform the public on matters of public interest, disclosure is also required unless the content has undergone genuine human editorial review. The European Commission published a first draft Code of Practice on AI content marking in December 2025, with full enforcement of the transparency obligations set to apply from August 2026.

What Article 50 does not currently mandate is disclosure for every AI-generated advertising visual. An AI-produced background, a synthetic product render, a generated lifestyle image used in a display ad: none of these automatically trigger the disclosure requirement unless they imitate a real person or context in a misleading way. That grey zone is exactly what the Digiday framing captures. An AI face, a synthesised voice, a body that never existed: these sit at different points on a spectrum the industry has not drawn a consistent line across.

The FTC's position in the US is similarly incomplete on AI creative specifically. Its August 2024 rule banning fake AI-generated reviews addresses fabricated testimonials and synthetic endorsements. As we have reported on FTC compliance for affiliates in 2026, the emerging expectation is that affiliate program contracts should specify disclosure requirements for AI-assisted content and that program managers need to treat compliance as a structural issue in publisher onboarding, not an afterthought.

The disclosure question Google has already answered differently

There is one additional wrinkle specific to content publishers. Google's Search Quality Evaluator Guidelines treat blanket AI disclaimers as a credibility signal in the wrong direction. As we examined when looking at Google's stance on AI content in affiliate marketing, a site-wide disclaimer stating that some content may have been AI-generated is more likely to reduce perceived trustworthiness under E-E-A-T than to build it.

This creates a three-way tension for affiliates operating content sites with AI-generated advertising creative. Regulatory compliance may push toward disclosure. Search quality signals push against undifferentiated disclaimers. And the NYU/Emory research says the disclosure itself cuts performance. None of these pressures point in the same direction.

The answer is not to avoid disclosure. It is to be specific rather than sweeping. A label that says “this ad image was AI-generated” on the ad unit itself is different from a footer disclaimer that poisons the entire site's credibility. Our transparency guide for affiliate marketers makes the broader case: disclosure framed as a professional commitment lands differently than disclosure buried as legal cover.

What affiliate program managers should do now

The NYU/Emory study is a data point, not a directive. It measures CTR in a specific context, using specific formats, in a specific moment. Consumer attitudes toward AI-generated content are not static. As AI-produced visuals become the norm rather than the exception, the disclosure penalty may narrow.

What affiliate program managers cannot afford to do is treat the 31.5% figure as a reason to avoid the disclosure question altogether. The FTC is watching AI-generated endorsements closely, EU enforcement on AI content marking arrives in August 2026, and publishers whose audience relationships depend on credibility have more to lose from appearing evasive than from a short-term CTR dip.

The ad industry has not agreed on what honesty looks like in this context. For affiliates, that uncertainty is itself the risk. The publishers who define their own standard now, before regulators impose a coarser one, are in a better position than those waiting for external clarity that may arrive with penalties attached.