AI-powered fact-checking systems: accuracy, limitations, and influence on misinformation control

The rapid spread of misinformation on digital platforms threatens journalistic integrity and informed discourse, especially in culturally diverse nations like India. While AI-based fact-checking offers scalable alternatives to manual verification, its real-world performance remains limited. This study evaluates leading AI fact-checkers through a mixed-methods approach: benchmarking datasets (FEVER, LIAR, COVID-Fact) and qualitative analysis of 60 Indian claims. Results show that while models perform well on standard data, accuracy drops on regional or sociopolitical content due to retrieval errors, weak justifications, and poor multilingual support. Interpretability analysis reveals key architectural and data gaps. The above discoveries highlight the urgency of culturally based AI models, transparent reasoning, and technologist-journalist collaboration. The work is of great value in that it offers the much-needed direction in developing ethical and effective fact-checking tools, which will be applicable in the Indian media.