How can we fight misinformation in the AI era? Chinese media are testing new approaches — including The Paper's latest experiment using AI to support fact-checking.
ChatGPT, Gemini, DeepSeek and Doubao were recently put to the test. The results:
• AI is good at spotting claims already proven false
• Some models can even tell real photos from AI-generated ones
• But hallucinations persist, and models can still invent confident-but-false details.
The lesson? AI can assist verification — but the final judgment must stay with humans.
This trial is part of The Paper FactCheck's multi-year effort to improve China's fact-checking ecosystem. Since 2021, it has published 900+ fact-checks and built an open platform to strengthen public participation in verification — scan the QR code for more from the team.
If you're a journalist, which part of your fact-checking process would you trust AI with — and which parts never?
