This paper reports lexical entrainment in a multi-party human–robot interaction, wherein one robot and two humans serve as participants. Humans tend to use the same terms as their interlocutors while making conversation. This phenomenon is called lexical entrainment. In the field of human–robot interaction, lexical entrainment has been investigated as a one-to-one interaction, and it is still unknown how humans entrain to a robot and/or another human interlocutor in a multi-party interaction. In this study, we investigate which participant, a robot or a human, strongly entrains to the other human’s lexical choices in a multi-party group interaction. Moreover, we investigate whether witnessing interaction about whether a human is entrained to a robot affects the entrainment frequency of the other human participant. We conducted a map navigation task wherein a robot and two humans guide each other by describing icon images on the map. Our results showed that the human participants were lexically entrained to a greater extent to the robot than the human participant in the multi-party interaction. We found no significant effect proving that a human participant witnessing an interaction between a human and a robot would become more entrained to the robot or the other human participant.