Artificial intelligence has become a valuable tool and a major challenge for journalists. While AI can assist reporters with analyzing data, it has made it easier to create and spread misleading or false information. Open source verification methods can help to determine whether digital content is authentic, manipulated, or taken out of context.

During the Iran war, the spread of AI-generated and misleading visuals demonstrated how easily misinformation can circulate online. Videos claiming to show missile strikes and destruction were widely shared across platforms such as X and Telegram, but many were taken from unrelated events, pulled from video games, or entirely fabricated.

Some visuals were realistic enough to appear credible, while others mimicked satellite imagery in order to convey authenticity. As The New York Times noted, this made it difficult for viewers to distinguish real from fake content.

Dakota Flournoy, US editor of Storyful, shared with us her seven-point checklist for identifying AI-generated videos. This includes: analyzing logos, numbers and text for distortion or inconsistencies; looking out for unusual camera movement or objects behaving in ways that defy basic physics; identifying unnatural shadows or lighting; checking the duration (AI videos tend to be short or clipped together) and geolocating the footage.

As AI improves, it becomes all the more important for journalists to exercise sound judgment regarding the sourcing and context of the information. Remember to ask yourself the following questions:

However, not all misleading content is AI-generated. Real photographs or videos may be posted with captions claiming they were taken at a different time, location or event, deceiving viewers by presenting genuine images out of context.

Journalists also need to be careful when using AI tools themselves. As Rhona Tarrant of CBS News Confirmed observed, AI-generated information can be "really confidently wrong." To avoid presenting false conclusions, it is essential to verify all available information.