A collaboration between computer scientists and archaeologists at the University of Bologna has led to the discovery of four previously unknown archaeological sites in Iraq’s Abu Ghraib district. The findings, published in PLOS ONE, were made possible by an AI model trained to analyze historical satellite images from the U.S. CORONA program, a Cold War initiative that captured black-and-white photos of the Middle East in the 1960s. “Our deep learning model reached about 90% accuracy in detecting archaeological sites,” said Marco Roccetti, professor of Computer Science and Engineering and co-author of the study. “It even revealed new settlements invisible to the human eye, offering vital insights for heritage preservation in landscapes transformed by human activity.” The researchers focused on Abu Ghraib, near Baghdad, a region never systematically surveyed and heavily altered in recent decades. By training the AI on both modern high-resolution images and historical CORONA data, the system reconstructed a landscape now largely lost. “Half of the sites still visible fifty years ago are today almost destroyed,” explained Nicolò Marchetti, professor of Near Eastern Archaeology. The discovery demonstrates how AI can complement traditional fieldwork by rapidly screening large areas and pointing archaeologists toward promising locations. The team emphasizes that final validation remains in human hands, but technology accelerates research and opens new opportunities for studying vanishing cultural landscapes.