AI Error Makes NWS Map Invent Towns

Weather maps usually point you to places you can find on a real map. Over the weekend in Idaho's Camas Prairie, the National Weather Service shared one that highlighted winds over "Orangeotild" and "Whata Bod"—two towns that do not exist. "Hold on to your hats," a social media post said, predicting possible high winds in Orangeotild. The NWS confirmed the errors stemmed from an AI-generated base map used by its Missoula, Montana, office. The graphic, which also contained other misspellings and geographic mistakes, was posted on social media and deleted Monday after questions from the Washington Post.

"Recently, a local office used AI to create a base map to display forecast information, however the map inadvertently displayed illegible city names," said NWS spokeswoman Erica Grow Cei. "The map was quickly corrected and updated social media posts were distributed." It's not the first odd-looking forecast image from the agency, the Post reports. A November post from the Rapid City, South Dakota, office also featured misspelled locations and a visible Google Gemini logo, though NWS would not confirm if that one was AI-made.

The weather service says AI is being tested across the agency, from forecasting tools to graphics, but insists it evaluates accuracy and will stop using AI "where it is not effective." The glitch lands as the NWS and its parent agency, NOAA, lean harder into AI while dealing with staffing losses under the Trump administration's cuts. Former NWS official John Sokich said such experimental products are supposed to be clearly labeled, calling the lack of labeling likely "just an oversight."

Outside experts warn the stakes go beyond goofy town names. Missteps from a trusted government source, says Cornell misinformation scholar Claire Wardle, could chip away at confidence in both the agency and AI more broadly. Weather communicator Chris Gloninger notes this forecast was low-risk but worries similar errors in higher-impact situations could erode the public trust forecasters depend on. Research shows that some AI models can provide accurate 10-day forecasts, Gizmodo reports, but the models call for a human to confirm the information, which will be harder to do after NWS staff cuts.