Paper Number
ICIS2026-3129
Paper Type
Completed
Abstract
Traditional wisdom suggests that greater linguistic accuracy enhances the helpfulness of online reviews. With the rise of genAI, reviewers can now easily polish or even generate highly accurate text, making near-perfect language increasingly prevalent and often presumed to increase a review’s impact. Departing from this view, this study draws on attribution theory to propose that linguistic accuracy exhibits an inverted U-shaped relationship with perceived helpfulness in the genAI era. Using a dataset of 50,949 Yelp restaurant reviews collected after the release of ChatGPT, the results show that the accuracy of grammar, punctuation, and capitalization follows this inverted U-shaped pattern, whereas spelling does not. A follow-up experiment replicates the effect and uncovers the underlying mechanism: improvements from low to moderate levels of linguistic accuracy enhance helpfulness because they are attributed to human effort, whereas further increases from moderate to high levels reduce helpfulness because they trigger attributions of AI involvement.
Recommended Citation
Sun, Peiliang; Sun, Shuyu; and Liu, Jingwen, "When Perfect Language Backfires: The Impact of Linguistic Accuracy on Review Helpfulness" (2026). ICIS 2026 Proceedings. 27.
https://aisel.aisnet.org/icis2026/user_engage/user_engage/27
When Perfect Language Backfires: The Impact of Linguistic Accuracy on Review Helpfulness
Traditional wisdom suggests that greater linguistic accuracy enhances the helpfulness of online reviews. With the rise of genAI, reviewers can now easily polish or even generate highly accurate text, making near-perfect language increasingly prevalent and often presumed to increase a review’s impact. Departing from this view, this study draws on attribution theory to propose that linguistic accuracy exhibits an inverted U-shaped relationship with perceived helpfulness in the genAI era. Using a dataset of 50,949 Yelp restaurant reviews collected after the release of ChatGPT, the results show that the accuracy of grammar, punctuation, and capitalization follows this inverted U-shaped pattern, whereas spelling does not. A follow-up experiment replicates the effect and uncovers the underlying mechanism: improvements from low to moderate levels of linguistic accuracy enhance helpfulness because they are attributed to human effort, whereas further increases from moderate to high levels reduce helpfulness because they trigger attributions of AI involvement.
Comments
14_User_behavior