Mapping AI uses and perceptions in fact-checking organisations in Mediterranean Europe

Abstract

This study maps the uses and perceptions of AI among fact-checking organisations accredited by Duke Reporters' Lab, the International Fact-Checking Network, and the European Digital Media Observatory in France, Italy, Spain, Portugal, and Greece. These countries were selected as they are among the most populous territories in the Mediterranean European region and host the majority of fact-checking entities. A qualitative approach is used based on 20 semi-structured in-depth interviews with representatives from each organisation. The interviews were conducted between May and July 2024. The findings reveal that AI tools are primarily employed during the gathering, monitoring, and detection phase, where they help increase speed and expand the scope of disinformation tracking. There is also a growing use of Large Language Models (LLMs) as tools for compiling information and conducting preliminary searches. Generally, organisations interviewed report feeling adequately prepared to use AI, largely due to the accessibility and intuitive design of current tools. Nevertheless, factors such as lack of funding and job precarity, as well as staff composition, affect their implementation, with only seven organisations (36.8%) having developed proprietary tools. Although AI is viewed as a valuable resource for automating repetitive tasks such as transcription, translation, and monitoring, it also raises concerns. Organisations warn that AI could contribute to more sophisticated disinformation and even foster a Liar’s Dividend effect, undermining public trust and the credibility of fact-checkers themselves.