Signal Credible, Not Always Correct: How Reddit Users Verify AI-Generated Legal Advice

Summary

Amid persistent gaps in access to justice, laypeople are increasingly turning to large language models to resolve real legal problems. This paper presents evidence that the practical influence of AI-generated legal advice on platforms such as Reddit depends less on its accuracy than on the social production of credibility, the way users collectively vet, endorse and contextualize AI-generated answers. While prior research has focused on assessing the accuracy of legal AI outputs, this study instead examines the verification practices communities use, noting that plausible-looking but incorrect advice can still gain social traction. This raises access-to-justice and misinformation concerns about using legal AI in unregulated public forums. The researchers argue that community verification practices themselves deserve to become a new object of study.