How Generative AI Review Summaries Disrupt Users’ Evaluative Processes in Online Purchase Environments

Generative artificial intelligence review summaries (GARS) are increasingly integrated in online purchase environments to provide a condensed representation of peer-generated reviews. Although GARS offer an efficient means of conveying diverse peer-generated information, their juxtaposition with traditional representations—such as star ratings and peer reviews—may complicate users’ evaluative processes. Specifically, when GARS conflict with traditional representations, e.g., a positive GARS alongside a negative star rating, they may induce ambivalence, a state characterized by concurrent positive and negative evaluations, increasing the difficulty of purchase decisions. We hypothesize how this may spur an increased gaze transition entropy (GTE)—a measure which captures the volatility of eye movements across different areas of interest and the efficiency of visual attention—leading to less efficient information processing during product evaluation. To test our hypotheses, we conducted a within-subjects lab experiment employing eye tracking to capture gaze behavior in response to varying states of alignment between GARS and traditional representations. Our results confirm that misalignment between GARS and traditional representations significantly increases both purchase difficulty and GTE, and that purchase difficulty significantly mediates such misalignment’s impact on GTE, indicating how users may experience reduced visual efficiency due to misalignment. These findings carry implications for research and practice related to online purchasing, suggesting how misalignments between GARS and traditional representations can disrupt decision-making in online environments.