Visual AI in Advertising: 19% More Clicks with GenAI-Generated Ads
A groundbreaking study by researchers from NYU Stern and Emory University provides the first empirical evidence: fully AI-generated advertisements significantly outperform human-created creations. However, the results also reveal surprising limitations and a critical dilemma regarding AI disclosure that is particularly relevant for European companies under the EU AI Act.
Key Findings at a Glance
The study "The Impact of Visual Generative AI on Advertising Effectiveness" by Lee, Todri, Adamopoulos, and Ghose (2025) examined three approaches: human-created ads, AI-modified ads, and fully AI-generated ads. The results have far-reaching implications for your marketing strategy.
Research Design: How the Study Was Conducted
The researchers combined laboratory experiments with an extensive field study on Google Ads to examine both the psychological mechanisms and real-world market performance. The data comes from an international beauty retailer with actual campaign data.
Methodological Approach
- Laboratory experiment: 685 participants at two US universities evaluated purchase intention and ad effectiveness
- Field study: 4 months on Google Display Network with 105,999 impressions and 4,026 clicks
- Three ad types: Human-created, AI-modified (5 variants), AI-generated (4 variants)
- Visual GenAI tools: Stable Diffusion, Midjourney, and DALL-E compared
Finding 1: Creation Beats Modification
The most surprising result of the study: AI-modified ads bring no improvement over human-created ads. On the contrary, they perform even worse. Only when AI creates entirely new ads from scratch does its superiority become apparent.
The reason lies in the concept of output constraints : when modifying, AI must work within predefined visual boundaries. These limitations restrict its creative capacity. With complete new creation, AI can fully exploit its generative potential.
Finding 2: Why AI-Generated Ads Work Better
The researchers identified three main mechanisms that explain the superiority of AI-generated ads. These insights help you understand how to optimally deploy Visual GenAI.
AI-generated ads produce significantly stronger emotional responses (β = 0.23). They create coherent visual compositions that emotionally engage viewers and enhance both informative and persuasive effects.
AI-generated ads achieve higher aesthetic quality (β = 0.13) and more balanced product-background relationships. This reduces cognitive load and facilitates information processing.
AI-modified ads lose ecological validity (β = -0.52), appearing less authentic. AI-generated ads avoid this problem as they emerge as a coherent unit.
Interestingly, both AI approaches show similarly high values for perceived novelty . The difference lies not in originality but in the appropriateness of the creative execution.
Finding 3: The AI Disclosure Dilemma
Perhaps the most critical finding for European companies: disclosing AI use drastically reduces advertising effectiveness. This directly conflicts with the transparency requirements of the EU AI Act.
Regulatory Context
- EU AI Act (Article 50): Mandates disclosure of AI-generated or manipulated content
- US Executive Order: Calls for effective labeling and provenance mechanisms for AI content
- Trade-off: Compliance vs. performance requires strategic consideration
- Recommendation: Develop flexible disclosure strategies that balance transparency and effectiveness
Implications for the European Market
For European companies, this study presents particular challenges and opportunities. The EU AI Act, GDPR, and traditionally high quality expectations of European consumers require a differentiated approach.
Regulatory Requirements
EU AI Act Compliance for Advertising
- Article 50: Disclosure requirement for AI-generated image, audio, and video content
- GDPR compliance: Data protection in AI training and personalization
- Consumer rights: Ensure transparency about AI use in advertising
- Documentation: Guarantee traceability of AI usage
Opportunities for European Companies
SMEs in particular can create professional advertising materials through Visual GenAI without expensive agencies. The study shows: AI-generated ads are not only cheaper but also more effective.
A/B testing with numerous variants becomes practical. European companies can test more ad variants and optimize based on data.
The combination of AI creation and human fine-tuning aligns with European quality standards. GenAI for ideation, experts for finalization.
Early development of disclosure strategies can become a competitive advantage when competitors must catch up later.
Challenges for the European Market
The disclosure problem hits European companies particularly hard, as the EU AI Act imposes strict transparency requirements. At the same time, European consumers are traditionally more skeptical of AI-generated content.
Success Factors for European Companies
- Hybrid workflows: AI for creation, humans for modification and compliance
- Transparent communication: Develop proactive disclosure strategies
- Quality control: Human review of all AI-generated content
- Data protection: Establish GDPR-compliant AI tools and processes
The study clearly shows: the key is not to avoid AI but to deploy it strategically. European companies that establish the right processes now will benefit in the long term.
Practical Implementation: How to Optimally Use Visual GenAI
Based on the study results, clear recommendations can be derived. The key lies in the right division of labor between AI and human expertise.
Phase 1: Ideation with GenAI
Use Visual GenAI in the early creative phase to quickly generate many ad concepts. Let AI create completely new designs rather than modifying existing ones. Test different prompts and styles to explore the range of possibilities.
Phase 2: Selection and Optimization
Have marketing experts select the best AI-generated concepts. Conduct A/B tests to identify the most effective variants. The study shows: even AI-generated ads benefit from professional pre-selection.
Phase 3: Human Finalization
Leave final adjustments to human experts. Logo placement, brand consistency, and last refinements require the contextual understanding and aesthetic judgment that humans excel at.
Best Practices from the Study
- Prefer complete new creation: Let GenAI create entire ads, not just modify parts
- Include product packaging: AI-generated packaging designs additionally enhance the positive effect
- Test multiple tools: Stable Diffusion, Midjourney, and DALL-E deliver different results
- Plan disclosure strategically: Develop communication strategies that balance transparency and effectiveness
Text AI vs. Visual AI: Different Strengths
A fascinating result of the study: Visual GenAI behaves fundamentally differently from text AI (LLMs). While LLMs perform better at editing existing texts, it's exactly the opposite for visual AI.
Better at editing and refining existing texts. Grammatical and semantic rules structure the output. Chain-of-thought prompting improves results.
Better at complete new creation. No rigid rules, but free combination of visual dimensions. Higher creative freedom leads to better results.
Use LLMs as a "sounding board" for feedback and refinement of your texts. Let them optimize existing copy rather than write completely new ones.
Use Visual GenAI to create completely new ad images. Avoid only modifying existing designs.
Limitations and Outlook
Like any scientific study, this one has its limitations. For practical application, it's important to know these and interpret the results accordingly.
The study focuses on beauty products. Results could vary in other industries like fashion, electronics, or services. Further research is needed.
GenAI is evolving rapidly. The results reflect the state of 2024/2025. Future models might perform better at modifications too.
The study was primarily conducted in the US. European consumers might react differently to AI-generated advertising.
The study measures short-term effects (CTR, purchase intention). Long-term impacts on brand perception and customer loyalty were not examined.
Conclusion: Using Visual GenAI Correctly
The study provides clear evidence: Visual GenAI can significantly increase advertising effectiveness, but only if you use it correctly. The key lies in complete new creation, not modification of existing ads.
Key Takeaways
- Creation over modification: AI-generated ads outperform human ones by 19% CTR, AI-modified bring no improvement
- Consider disclosure dilemma: AI disclosure reduces effectiveness by 31.5%, but EU AI Act requires transparency
- Establish hybrid workflows: GenAI for ideation, humans for finalization and compliance
- Understand mechanisms: Emotional engagement, visual fluency, and ecological validity drive success
For European companies, this means: now is the right time to strategically integrate Visual GenAI into your marketing processes. The technology is mature, the benefits are measurable, and those who establish the right workflows early will benefit long-term. At the same time, the regulatory framework of the EU AI Act requires a thoughtful disclosure strategy that balances transparency and effectiveness.