Description
This project examines the risks of using AI chatbots—particularly Grok on X—as real-time fact-checking tools in social media environments, where they may generate or amplify misinformation through confident but inaccurate verification responses. It investigates how these systems produce misleading outputs under conditions of uncertainty and time pressure, and how users interpret and trust these answers, including the role of hallucinations in shaping beliefs. Combining journalistic analysis by Maldita.es with experimental behavioural research at the University of Jaén, the study documents real cases of Grok’s verification failures, identifies recurring patterns and triggers, and tests how exposure to such responses affects user trust, cognition, and behaviour. It also explores how interaction dynamics like public prompting influence AI outputs, how reliance on these tools may shift trust away from traditional authoritative sources toward AI-generated answers, and how cognitive biases such as confirmation bias and the continued influence effect make users more vulnerable to misinformation, ultimately aiming to understand how AI is reshaping verification practices and its implications for information integrity and public trust.
“Grok always seems confident in its answers; there are never any shades of uncertainty. If you ask whether something is real or not, it will take a clear position, and in many cases it will be wrong. Through the “Grok, is this real?” prompt, we want to study this phenomenon and examine how relying on an AI like Grok can contribute to the spread of misinformation. We also want to investigate whether Grok’s parameters and behavior when answering this type of question have changed over time, and what factors its responses are based on.” — Patricia Rojas, project coordinator
Project Team
Media outlet
Fundación Maldita.es contra la desinformacion: periodismo, educacion, investigacion y datos en nuevos formatos
Madrid, Spain
https://maldita.es
Coral García
Investigations Coordinator
Spain
Research organisation
