There is no question that being able to judge the reliability of information sources plays a crucial role in knowledge acquisition. In acquiring knowledge, one must be able to evaluate the sources of the information used. Much of what people know they learn from others, so they must be able to assess the trustworthiness of others. Human societies are built on trust. This paper focuses on cases of knowledge transfer between artificial systems (social robots) and human agents where language plays a role. After a brief presentation of strategies for assessing the trustworthiness of human-generated content, the question is raised as to what requirements artificial systems must meet in order to be considered trustworthy agents in a knowledge transfer. To this end, this paper will show that the application of large language model technology challenges many of our techniques for trustworthiness assessment. This means that social robots equipped with this technology have the potential to contribute to an epistemological crisis and undermine trust. Therefore, it is important to think about how to address this crisis. In order to contribute to “culturally sustainable social robotics,” issues of transparency, trustworthiness, reliability, and authenticity in the construction of social robots will be discussed.