The credibility of information generated by AI chatbots: an analysis of online discussions in Reddit

Introduction. The present investigation elaborates how participants of online discussion in Reddit assess the credibility of information generated by AI chatbots. Method. The empirical findings draw on the descriptive quantitative analysis and qualitative content analysis of 1464 posts submitted by 919 individual participants to 30 Reddit discussion threads during the period of October 2023 – September 2025. Analysis. The analysis focuses on the criteria by which the online participants assessed the credibility of information generated by AI chatbots. Second, it was examined how the credibility assessments were grounded by making use of various answer types. Results. The most frequently used credibility criteria were correctness, trustworthiness and usefulness of information, as well as chatbot´s capability to generate relevant information. Accuracy, coverage, currency and verifiability of information were employed less frequently. Overall, the findings indicate that online participants adopted a critical stance on the credibility of information generated by AI chatbots. In particular, the occurrence of hallucinations undermined the beliefs that chatbots offer correct and trustworthy information. The credibility assessments were mainly grounded by drawing on personal opinions and experiences obtained from the use of chatbots. To a lesser extent, the assessments were also grounded by drawing on comparison and explanation. Conclusion. Although the credibility of information generated chatbots is evaluated critically, people are not expecting that AI tools would offer fully credible answers.