Reporter’s Guide to Detecting AI-Generated Content

The guide addresses the growing problem of AI-generated misinformation. The guide highlights the shift in challenges for journalists. While it once took time to create false content, AI can now produce it in minutes, making traditional fact-checking methods obsolete. Relying on outdated detection techniques from even a year ago could be more dangerous than admitting uncertainty, as AI models are constantly improving and can easily bypass them.

To combat this, the guide presents a new, multi-faceted approach. It introduces seven advanced categories for journalists to master and a three-tiered system for verification. This system allows for different levels of scrutiny depending on the situation: a quick 30-second check for breaking news, a five-minute technical verification for standard stories, and a deep investigation for high-stakes reporting.

Key detection methods are detailed, including looking for "anatomical and object failures" where things appear "too good to be true," "geometric physics violations" where AI fails to adhere to natural laws of perspective and shadows, and "voice & audio artifacts" which are subtle tells in synthetic speech.

Ultimately, perfect detection may be impossible. The focus has shifted from finding definitive proof to a probability assessment and informed editorial judgement, urging journalists to adapt as they have to previous technological changes.