Language models have become an essential part of the burgeoning field of AI Psychology. I discuss 14 methodological considerations that can help design more robust, generalizable studies evaluating the cognitive abilities of language-based AI systems, as well as to accurately interpret the results of these studies.