This chapter presents an innovative computational method for detecting possible deception in the provenance texts of artworks. This is important because false ownership histories can conceal the kind of textual “red flags” that alert experts to potentially looted art. The first section presents the study’s scope, context and approach to automating the detection of Nazi-looted art by analysing provenance texts and explains the urgency of solving the problem of deception. The second part presents provenance datasets, and a digital tool developed for counting words linked to unreliability, uncertainty or anonymity, which may indicate potentially deceptive language. The third part analyses the strengths and weaknesses of this methodology, exploring lessons learned. Benefits of the methodology described here include transparency, objectivity and replicability, as well as the generation of quantitative and Boolean metadata for artworks that can be fed into ranking applications, recommender systems, knowledge graphs, and AI, in order to potentially quantify deception within the art market. Challenges include optimization of ranking methods and the tension between open data and closed archives, which complicates the verification of the artworks flagged. The chapter concludes with opportunities for future developments in different domains.