We examine how open and closed review posting policies play differentiating roles in creating social media bias. As a supplementary method to existing ones detecting fake reviews, we develop a trust measure estimating how genuine the review is, based on the frequent usage of strongly positive or negative words. Using the hotel industry as our application context, we empirically demonstrate that our trust measure serves as a correction factor that reduces social media bias. Interestingly, we observe particular hotel service features revealing strong upward manipulation to promote the businesses (for example, positive overall recommendation, interesting surroundings, and personal travel). By contrast, we identify some other features that reveal the presence of strong downward manipulation (for example, negative overall recommendation, disappointing room amenities, and poor atmosphere). From a practical perspective, this research can help both managers and consumers make better informed decisions by reducing the impact attributable to social media manipulation. •Our trust measure based on extreme terms supplements existing algorithms reducing social media bias.•We empirically support the validity of our extreme-term-based approach.•We identify particular hotel service features vulnerable to upward or downward manipulation.•Overall positive and negative evaluations are reduced after we correct for fake reviews.