Enabling and Understanding Personalization in AI-Generated Advertising Imagery
| Source: arXiv AI
Tags: personalization, AI advertising, image generation, consumer behavior, generative AI
A controlled study with 100 participants finds AI-generated advertising images perform best at moderate personalization — high personalization boosts perceived relevance but triggers a creepiness effect that overrides positive intent signals across all three measured outcomes.
Details
Researchers from Germany built an AI framework that generates personalized advertising images directly from customer data, then tested it in a controlled within-subject study with N=100 participants across four product types and three personalization levels. Participants rated each ad on attitude toward the advertisement, attitude toward the product, and purchase intention. The central finding: moderate personalization outperforms both no personalization and maximum personalization. At high personalization, participants correctly perceived the ads as more tailored — but this also triggered higher perceived creepiness, which had a stronger negative effect on all outcomes than the personalization signal had a positive one. The results carry direct implications for companies deploying generative AI in marketing workflows. More data and more personalization is not always better; there appears to be a sweet spot where AI-generated imagery feels relevant rather than surveillance-adjacent. The study tests conditions beyond simple text swaps, examining full image generation conditioned on customer profiles. A companion paper (AdMan, arXiv:2609.12694) by overlapping authors describes the technical pipeline behind this study, providing both the infrastructure and the consumer validation together.