Konsumentenwahrnehmung KI-generierter Produktbilder im E-Commerce

Abstract

Generative AI enables the rapid and cost-effective generation of photorealistic product images, fundamentally changing the creation of visual content in e-commerce. This opens up opportunities for efficiency and innovation but raises questions about trust and authenticity, especially when consumers cannot reliably identify the origin (AI vs. human) or when AI labeling influences their perception. At the same time, the EU Artificial Intelligence Act will introduce a requirement for visible labeling of authentic-looking AI-generated or AImanipulated images as of August 2, 2026. For providers, this raises the question of whether and to what extent such AI labeling changes the perception of their product images. First, a theoretical framework of perception, trust and acceptance research is provided. From this, three central evaluation criteria were derived, which were used to assess the stimuli used in the empirical study: image aesthetics, product quality and image credibility. This is followed by an introduction to the technological basics of AI image generation and its potential for product image creation, as well as a discussion of the relevant legal framework. In an online survey study the influence of AI labeling, the actual image source (AI vs. human) and the perceived image source (AI vs. human) on the evaluation of product images were empirically examined. In addition, the ability to recognize AI-generated images as such was tested and attitudes toward AI labeling and labeling requirements were surveyed. The results of the empirical study show that visible AI labeling has a significant negative impact on the aesthetic evaluation of product images, while no significant effect was observed on product quality and image credibility. Without labeling, there is no significant difference in the ratings between AI-generated and human-created product images. Instead, the perceived source is decisive. Product images that participants classified as human received significantly higher ratings for product quality and especially image credibility. The recognizability of AI-generated product images without labeling is image-specific but overall at a random level. In addition, a clear majority is in favor of mandatory labeling of AI-generated product images in e-commerce.

Date of Award	2025
Original language	German (Austria)
Awarding Institution	Johannes Kepler University Linz
Supervisor	Harald Kindermann (Supervisor)