Archive · 21 Nov 2025
AI Chatbots Fail to Detect Own AI-Generated Images
AI chatbots, used to check if pictures are real or fake, sometimes can't tell if the pictures they made themselves are real.
This is a problem because people use these chatbots to check if pictures are real.
Experts say we still need people to check if pictures are real, not just AI.
AI-powered chatbots struggle to verify authenticity of images, including those generated by themselves
This raises concerns about their reliability in real-time fact-checking
AI tools often fail to identify images as fake, even when generated by the same model
Examples include a fabricated image of a Philippine lawmaker and a fake protest image in Pakistan-administered Kashmir
Experts warn that AI cannot replace human fact-checkers in combating AI-generated misinformation
- Who
- AI-powered chatbots, users sharing fake images
- What
- Failed to detect authenticity of AI-generated images
- When
- Currently, with increasing use of AI tools for real-time fact-checking
- Where
- Online platforms, including social media
- Why
- Due to limitations in AI models' visual understanding and mimicking capabilities
AI Proponents
AI Skeptics
AI's Potential in Fact-Checking
AI Proponents
AI can help quickly geolocate images and spot visual clues for authenticity
AI Skeptics
AI cannot replace human fact-checkers due to limitations in visual understanding and mimicking capabilities
- AI Model
- Google's Gemini, Microsoft's Copilot, Nano Banana
- Image Circulation
- Over a million views across social media
- Fact-Checking Challenges
- Hyperpolarized societies, bias accusations, scaling back human fact-checking
- AI's Role
- Useful for quick geolocation and spotting visual clues, but cannot replace human fact-checkers
- Research Findings
- Columbia University study: 7 AI chatbots failed to correctly identify image provenance