For the past couple of years, the advertising industry has buzzed with talk of generative AI. Tools like DALL-E and Midjourney promise to revolutionize creative work, but much of the conversation has stayed stuck in vague promises and flashy demos. What marketers really need is hard evidence on whether these technologies deliver in the real world. A fresh study from October 2025, led by researchers Lee, Todri, Adamopoulos, and Ghose, finally provides that. Their findings cut through the noise: generative AI can boost ad performance, but only under specific conditions, and it demands skilled human oversight to shine.
The research dives into visual advertising with a rigorous setup that mirrors everyday campaigns. The team analyzed over 100,000 ad impressions on Google Ads, using actual budgets and target audiences. They pitted three main types of creatives against each other: traditional human-designed ads as the baseline, human designs lightly tweaked by AI, and entirely AI-generated visuals from models like Stable Diffusion or Midjourney. They also tested a twist, where AI handled not just the ad but the product packaging too.
The numbers tell a compelling story. Fully AI-created ads saw a 19 percent jump in click-through rates compared to human ones. AI-modified versions, however, showed no improvement at all. When AI extended to packaging, performance climbed another 15 percent. But here’s the twist: simply labeling an ad as AI-generated slashed click-through rates by about 31 percent. These results confirm something counterintuitive. AI thrives as a bold idea generator early in the process, not as a subtle editor later on. For marketers chasing efficiency, this shifts the playbook from cautious tweaks to unleashing creativity from the start.
So why do completely fresh AI designs outperform the rest? It boils down to how these tools handle freedom versus restrictions. When AI builds from scratch, it avoids the rigid rules of existing layouts, colors, or fonts that cramp its style. This liberty unlocks real advantages in how audiences respond. First, AI images often feel smoother to process. Viewers’ eyes glide over them more easily, creating a quick sense of familiarity without effort. Second, they spark deeper emotional connections, pulling people in with unexpected visuals that resonate on a gut level. Third, when AI designs the whole package, from ad to product visuals, everything aligns seamlessly. The result is a cohesive look that feels polished and intentional, like a brand speaking in one clear voice. In essence, AI excels at weaving complex elements into harmony right from the outset, rather than patching up human drafts after the fact.
Yet the study uncovers a thorny challenge: transparency comes at a cost. Regulations in places like the EU and US now require disclosing AI involvement, but revealing it bluntly tanks engagement. The drop isn’t about ethics; it’s psychology at play. People appreciate the sleek appeal of AI aesthetics, yet a stark “AI-generated” tag stirs doubts about authenticity. It whispers that the work lacks a human touch, eroding trust just when curiosity peaks. Marketers face a tightrope here. Full honesty matters for compliance, but it can undermine results.
The fix lies in smarter communication. Instead of a blunt stamp, try subtle framing like “enhanced with AI creativity” or “powered by innovative tools.” Place it in context, perhaps in a footnote or story element, rather than front and center. Testing shows these approaches preserve much of the lift while meeting legal needs. Over time, as audiences grow comfortable with AI, this backlash may fade, especially among younger demographics who already shrug off such labels. For now, though, it underscores that words matter as much as images in this space.
What does all this mean for marketing teams? The study doesn’t herald the end of human creatives; it maps a smarter collaboration. Start by deploying AI in the brainstorming phase. Let it flood the table with mood boards, concept variants, and wild ideas, where variety fuels breakthroughs. Humans then step in to curate, sifting through the volume to pick winners. This flips the creative dynamic from solo crafting to expert selection. Don’t overlook packaging either, as integrated AI design builds instant credibility. And always measure beyond clicks: track conversions, brand sentiment, and sales to see the full picture. Success isn’t in the tech alone; it’s in granting AI the space to invent, not just polish.
Of course, no study is flawless, and this one has clear boundaries worth noting. The experiments focused on beauty industry campaigns, a field obsessed with visual polish and emotion. Results might not translate as cleanly to drier sectors like finance or B2B services, where trust and facts often trump aesthetics. The AI was wielded by pros who mastered prompts and fine-tuning, a far cry from rushed attempts by busy teams. In reality, outcomes hinge on user skill: AI amplifies talent, turning strong creatives into superstars while leaving novices with generic output. There’s also the risk of sameness, as these models draw from past hits, optimizing trends but rarely inventing them. This could lead to a bland visual landscape if everyone leans too hard on the same tools. Finally, the focus on click-through rates captures attention, not deeper impact like purchases or loyalty. Teams should dig further to validate real ROI.
In the end, this 2025 research points to a hybrid future for marketing: data-fueled, experimental, and profoundly collaborative. AI won’t eclipse human ingenuity; it will accelerate it, turning fleeting sparks into vivid realities in moments. Guided by sharp prompts, clear goals, and discerning eyes, it scales inspiration across teams and campaigns. For those ready to invest in expertise, the payoff is clear: not replacement, but a powerful new ally in the art of persuasion.
Further Reading
The Impact of Visual Generative AI on Advertising Effectiveness
This SSRN paper by Hyesoo Lee, Panagiotis Adamopoulos, Vilma Todri, and Anindya Ghose details the core experiment and findings on AI-created vs. human ads.
URL: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4927890
The AI Advertising Paradox
An NYU Stern overview of the study, highlighting the 31.5% CTR drop from AI disclosure and strategic implications for marketers.
URL: https://www.stern.nyu.edu/experience-stern/about/departments-centers-initiatives/academic-departments/marketing/faculty-directory/anindya-ghose/ai-advertising-paradox
Ethical Considerations and Disclosure of AI Use for Content Marketing
A practical guide from Convince & Convert on framing AI involvement to build trust while complying with regulations, including placement tips.
URL: https://www.convinceandconvert.com/ai/ethical-considerations-and-disclosure-of-ai-use-for-content-marketing/
Support the work. Shape the conversation. Donate today.