ISCAP Proceedings: Abstract Presentation
When a Review Sounds Like AI: The Effects of AI-Likeness of a Review on Helpfulness
Youngeui Kim
Appalachian State University
Namita Lamichhane
Appalachian State University
Abstract
Generative AI has lowered the cost of composing product reviews, expanding review supply while blurring the line between human- and AI-written content. This study raises a question about how consumers evaluate reviews that resemble AI-generated text but are attributed to human reviewers – we use the term "AI-likeness" since reviewers rarely disclose AI assistance outright. Rather than treating AI-likeness as a uniform negative cue, we draw on the Elaboration Likelihood Model (ELM) to examine how AI-likeness reshapes the persuasive power of the cues embedded within a review: information diagnosticity as a central route and source credibility as a peripheral route. We argue that AI-like features such as fluency and structural clarity ease the cognitive burden of processing diagnostic content (e.g., long reviews, diverse topics), strengthening its positive effect on helpfulness (H1), while the same features simultaneously undercut perceived authenticity, weakening the persuasive value of source-credibility cues (H2). We further examine that this trade-off tilts favorably for search goods (i.e., The AI-like reviews are more helpful for search goods), whose attributes are verifiable independent of firsthand experience, relative to experience goods (H3). Using 8,890 reviews, we find that AI-likeness significantly amplifies the effect of topical diversity, i.e., information diagnosticity, on review helpfulness, and it significantly weakens the positive influence of reviewer experience, a peripheral credibility cue. Consistent with H3, the helpfulness of AI-like reviews is significantly greater for search goods than for experience goods. These findings suggest that AI-like reviews are not necessarily perceived as less helpful but instead change how consumers rely on the central and peripheral routes of persuasion, contingent on product type. The study extends ELM research to account for AI-likeness as a route-specific moderator rather than a simple negative cue. We also offer review platforms an evidence base for designing AI-content filtering and display features.