Fake AI storm images overwhelm feeds, burying real Hurricane Melissa updates
Officials are urging the public to get news updated from verified sources only.
Images of Hurricane Melissa that were made with artificial intelligence (AI) software are inundating social media, overshadowing crucial storm reporting and safety information as the Category 5 storm threatens Jamaica with heavy rainfall and intense winds. This misinformation makes it difficult for the public to distinguish reliable updates, increasing confusion and jeopardizing safety during the storm.
Many of the videos are clearly watermarked with the Sora logo — OpenAI’s video generator tool, which went live on September 30.
Tiktok has removed at least two dozen fake videos so far, many of them depicting floods, destruction, and violent winds that are not consistent with on-the-ground, verifiable reports.
Other problematic AI-generated videos include supposed footage of people swimming and playing in the streets as a storm approaches, downplaying the storm’s significance. Many videos and images that have not been unpublished are now marked with community notes and labels referring to the AI sources.
Hard to tell real from fake
Some fake videos, like this aerial view said to be of Melissa, have been convincing enough to be shared again and again, with some users trying to pass it off as authentic:
And the AI phenomenon appears to have left people questioning the validity of REAL videos like this one from the US Air Force's ‘Hurricane Hunters.’ :
Jamaica’s information minister, Senator Dana Morris Dixon, told AFP she is aware of the fake images and is urging the public to listen to “official channels” for information updates.
The danger of using AI-generated images during a disaster
Experts say even seemingly innocent AI images can clog up social media channels, drowning out updates from verified sources.
“During emergencies, when people are stressed and need reliable information, such digital disinformation can cause significant harm by spreading confusion and panic,” reads a statement on York University’s website released in August 2025 in response to AI-generated images clogging up social media during B.C.’s wildfire season.
“This vulnerability to disinformation stems from people’s reliance on mental shortcuts during stressful times; this facilitates the spread and acceptance of disinformation. Content that is emotionally charged and sensational often captures more attention and is more frequently shared on social media.
Based on our research and experience on emergency response and management, AI-generated misinformation during emergencies can cause real damage by disrupting disaster response efforts.”
How to detect if an image or video is AI-generated
AI technology is improving at a dizzying pace, but if you look closely, there are still clues to look for when determining if a piece of content is real or computer-generated.
The Government of Canada's Get Cyber Safe has the following recommendations:
For images:
Look for strange, out of context details, like gibberish in place of writing, or building architecture that doesn't make sense.
Pay attention to missing body parts, especially hands.
Look for misplaced or "floating" objects.
Look for impossible perspectives, lighting, and shadows.
Overly-smooth texture is another telltale sign of AI.
Pay attention to the background. Is it overly smooth or blurred? This is another clue you could be looking at AI.
For video:
Look for language that is flat and monotone.
Speaking in short, choppy sentences is another tell.
Body language that is jerky or unnatural is a common sign of AI.
People often don't blink in AI videos.
You may also see shadows or sparks of light in places that don't make sense.
Some AI videos and images will contain a watermark or small logo indicating they are computer-generated.
Header image: Elements and AI-generated photo created using Canva Pro.