Imagine that the ad you see online was created not by a human designer, but by artificial intelligence. A study examines how visual generative AI (artificial intelligence that creates images) affects ad performance. The researchers used a combination of laboratory tests and real-world Google Ads campaigns to compare three types of ads: those created by human experts, those modified by AI, and those created entirely by AI.
When AI Fails and When It Wins
The research found that AI-modified ads, in which artificial intelligence merely enhances existing designs created by experts, do not perform better than the original human versions. In a laboratory experiment involving products such as shampoo or hand cream, consumers' purchase intent did not increase. Likewise, in a real-world Google Ads campaign that generated 105,999 impressions, these modified ads did not increase the click-through rate.
By contrast, ads created entirely by AI outperformed both other variants. In the field study, they increased the click-through rate by as much as 19%, representing an increase of 0.71 percentage points compared with human-created ads. This was evident for products such as men's cosmetics and body oil. Ads for which AI also designed the product packaging produced even better results, as this imposed the fewest constraints on AI creativity.
The success stems from the fact that AI-created ads evoke stronger emotional engagement among viewers and make visual information easier to process. This contrasts with modified versions, which often lose their naturalness and ecological validity. The research used models such as latent diffusion models to create ads and analyzed participants' responses in the laboratory.
Disclosing AI Is a Major Problem for Success
If an ad discloses that it was created or modified by AI, its effectiveness drops by as much as 31.5%, representing a decline of 1.17 percentage points in the click-through rate. This is important in the context of new regulations, such as the EU AI Act of 2024, which requires AI-generated content to be labeled. The study tested this in real-world campaigns using labels such as "AI-designed" or "AI-generated".
Digital advertising spending reached approximately USD 469.8 billion in 2024 (about CZK 10.805 trillion), and generative AI could significantly reduce this amount. The authors emphasize that AI is best used for complete creation, not merely for modifications.
