Solving Historical Ciphers with AI: Analysis of GPT’s Capability in Processing and Deciphering Cryptographic Postcards

Abstract

In this case study, we focus on the capabilities of ChatGPT for assisting in the research of historical encrypted documents. Our study is based on the dataset of historical encrypted postcards, typically with simple (substitution) ciphers. We split the task into two parts: transcription and interpretation of the postcard image, and then decipherment of the transcribed text. We summarize the current capabilities of (multimodal) AI tools in general, and specifically of GPT-5.2 / GPT-4o that were used as AI tools in our analysis. Following our cryptanalytic pipeline, we show specifically how the AI tool was capable of solving selected typical exemplars from our dataset. We show that GPT-5.2 is already capable of solving most of the required tasks with acceptable quality, and can be useful even for researchers not specifically trained in prompt engineering or artificial intelligence usage.