Instead of establishing trust through defining compliance-based standards like protocols augmented by cryptographic methods, it is shown that trust can emerge as a self-organizing phenomenon in a complex dynamical system. It is assumed that trust can be modeled on the basis of an intrinsic property called trustworthiness in every individual i. Trustworthiness is an objective measure for other individuals , whether it is desirable to engage in an interaction with i or not. Trustworthiness cannot directly be perceived. Building trust, therefore, relates to estimating trustworthiness. Subjective criteria like outer appearance are important for building trust as they allow the handling of unknown agents for whom data from previous interactions do not exist. Here, trustworthiness is grounded in the strategies of agents who engage in an extended version of the iterated prisoner's dilemma. Trust is represented as a preference to be grouped together with agents with a certain label to play a game. It is shown that stable relations of trust can emerge and that the coevolution of trust boosts the evolution of cooperation.