Web conversational services are exposed to several threats in which the social context between communicating participants is manipulated. Cybercrimes based on identity misrepresentation to obtain sensitive information are on the rise. Various scams and frauds are conducted by distributing malicious content, viruses, and spam over established communication sessions. In order to maintain overall security and enhance privacy, methods of estimating trustworthiness and reputation should be built into Web calling services. In this paper, we propose "TrustCall" a reputation-based trust model for real-time Web conversational services. In our approach, the reputation of a caller is evaluated using Authenticity Trust and Behavioral Trust. Authenticity Trust describes the legitimacy of a caller by collecting recommendations from other members of the network, whereas Behavioral Trust determines a caller's popularity based on its communication behavior. Other contributions include a threat taxonomy for Web calling services, including social threats, that directly target users. A set of experiments are conducted in order to prove the feasibility and effectiveness of our model