Fan Liu, Jiaming Liu
Online reviews are widely used to evaluate tourism performance, but it remains unclear whether platform ratings adequately reflect the underlying tourist experience. This study uses 67,744 cleaned Ctrip reviews from 112 A-level scenic spots in Liaoning Province, China, to examine what online reviews reveal beyond conventional satisfaction metrics. The final analytical sample comprises 106 threshold-qualified attractions with at least 100 reviews, supplemented by six highly reviewed sub-attractions that were listed separately on the platform but belonged to officially recognized A-level scenic systems. We combine topic modelling, sentiment analysis, and a rating–sentiment analytical framework to identify experiential dimensions, emotional patterns, and attraction-level sentiment risk. The results reveal a five-dimensional structure of tourist experience, including accessibility and ticketing, natural landscape imagery, cultural heritage interpretation, service-process quality, and overall affective appraisal. Positive sentiment is concentrated in landscape, heritage, and holistic appraisal themes, whereas negative sentiment is more prominent in accessibility and service-process dimensions. Quadrant-based analysis further shows that favourable ratings may coexist with relatively negative textual sentiment, suggesting that platform ratings and review-text sentiment do not fully converge. To extend review-level evidence to the attraction level, the study develops an attraction-level sentiment-risk indicator that captures the concentration of sentiment-negative reviews within each scenic spot. The findings suggest that online reviews function as a dual-channel evaluative system and can support sustainable destination management through more sensitive monitoring of operational friction and experiential risk.