2 AI-generated images circulated as flood scenes from Chattogram

All captions and news report excerpts in this article have been translated from Bangla into English

Two aerial images claiming to show the entire city of Chattogram submerged in floodwaters have recently been circulating across multiple social media platforms. A fact-check by Dismislab found that both images were created or edited using Google’s AI tools.

On July 10, the images first came to Dismislab’s attention through a post published by a verified Facebook page called “Shikkha Barta” (meaning “Education News”). The caption read, “Remember — when Chattogram, Noakhali, Feni, Cumilla, Sylhet and Sunamganj were flooded, the people of Chattogram rushed to help them. Today, all of Chattogram is going under water, yet hardly anyone is paying attention. Major media outlets are also not giving it adequate coverage, and there is no notable discussion in civil society either. Remember, Chattogram is an important and self-sufficient region of Bangladesh in every respect. This darkness will pass and light will come very soon, inshallah. If any part of the country falls into trouble, remember us again. Just as before, we will rush to stand by humanity. All communities from Chattogram — including those from the University of Dhaka and the University of Chittagong — please come forward from your respective positions. May Allah protect my beloved city of Chattogram. Amin. Collected.”

The same images were also posted across multiple other Facebook accounts ( 1, 2, 3 ) and spread individually and on YouTube, Instagram and X, formerly Twitter, as well.

The first image shows an aerial view of a vast city submerged in floodwater in bright daylight. Multi-story buildings and green trees in the city appear partially inundated. In the middle of the image, a semicircular flyover or bridge is visible, with none of it submerged in water. 

A second image of a similar nature shows an aerial view of a flooded city. A long and winding flyover or highway runs through the middle of this image as well. In this version, a portion of the flyover is submerged. The city’s roads and the ground floors of most buildings are underwater. The lower section of the image shows large amounts of floating debris scattered across the water, and in the lower right, a large vessel is seen crowded with people.

To verify the images, Dismislab examined them closely. In one image, several buildings in the middle of the water appear to blur and merge indistinguishably into the floodwater — their rooftops blend with the color of the floodwater, even though the front faces of the buildings remain visible.

In the other image, a flyover is shown with one end sloping downward into the water and extending toward a multi-story building. Several boat- and ship-shaped objects are also visible in the lower section of the image. The presence of such large vessels in an urban setting would be highly unusual.

For further verification, Dismislab tested the images using Google Gemini’s AI-detection tool “Verify AI.” The tool confirmed that both images were created or edited with the assistance of Google’s AI tools. Subsequent analysis of the Google SynthID heatmap confirmed that nearly the entire content of one image was AI-generated, while the middle and lower sections of the other image had been edited using AI.

In conclusion, the two aerial images being circulated with the claim that all of Chattogram is submerging in floodwater do not depict any real scene and are AI-generated.

It should be noted, however, that a severe flood situation has developed in the Chattogram region since July 4, due to extremely heavy rainfall and hill torrents. According to a report published by Prothom Alo on July 12, seven districts and 58 sub-districts across Bangladesh had been inundated by flooding as of 6pm on July 11. The death toll from flooding, hill torrents and landslides triggered by the extreme rainfall had reached 44. The number of injured stood at 39. In addition, 267,918 families had been trapped by floodwater, and 1,022,963 people had been affected.