Unfamiliar face matching is a difficult task. In typical experiments, viewers see isolated face pairs and have to decide whether they show the same or different people. Recent research shows that embedding faces into passports introduces a response bias, such that viewers are more likely to accept two pictures as showing the same person. Here, we investigate the cause of this bias. In a series of experiments, we vary the apparent authority of the identity documents, testing passports, driving licences, and student ID. By comparison to isolated face matching, the results show a bias towards responding same person for each document type. However, when ID information (name, date of birth, etc.) was removed from documents, the induced bias disappeared. We conclude that bias does not rely on perceived authority, but instead seems to occur only in the presence of identifying information, despite that being task irrelevant.