Google Gemini Partners With Watch Duty to Clear Static From Firefighter Radio Google's Gemini team, working with wildfire-tracking app Watch Duty, cut radio transcription word error rate by about 15% and raised volunteer satisfaction with transcription tools from roughly 35% to 92%, the company announced Friday. The tool, trained on more than 20,000 hours of Watch Duty radio recordings, transcribed a static-filled 911 call in under 30 seconds during Washington state's Railroad Fire in July 2026, enabling Watch Duty's first-ever nighttime Critical Alert. The project was completed at no cost through a Google.org fellowship, and the resulting Watch Duty features are free to use. On Friday, Google’s Gemini team announced the results https://blog.google/company-news/outreach-and-initiatives/google-org/watch-duty-gemini-tool/ of a new partnership with the wildfire-tracking app Watch Duty https://app.watchduty.org/ , after Gemini engineers spent six months embedded with Watch Duty’s first responders https://www.cnet.com/home/security/home-and-wildfire-safety-combined-rings-watch-duty-partnership-and-what-it-means/ . The results cut vital new wildfire dispatch translations down to seconds and helped improve radio communications https://www.cnet.com/tech/mobile/best-phones-under-300/ , which can be rough on the front lines of a new fire https://www.cnet.com/home/security/never-make-these-mistakes-to-if-your-home-is-threatened-by-fire/ . The Gemini AI improvements got a real-world test during Washington state’s Railroad Fire in July 2026 https://app.watchduty.org/i/110478 , when Watch Duty issued what it calls its first-ever nighttime Critical Alert. The alert stemmed from a nearly unintelligible, static-filled 911 call that the Gemini tool transcribed in under 30 seconds. That kind of assistance can help Watch Duty send alerts to affected communities more quickly and enable first responders to update maps more accurately and make better-informed decisions. “This is something the tool also has the capability to do if reports of flooding are communicated via radio https://www.cnet.com/home/security/popular-wildfire-app-watch-duty-expands-to-cover-nationwide-floods/ ,” said Dave Merritt, co-founder and chief technology officer of Watch Duty. “Across disasters, we hear feedback every day that first responders are using Watch Duty for additional situational awareness, especially for disasters that span multiple counties.” More from CNET Watch Duty reporters are active and retired first responders who can manually monitor up to 20 different feeds, though static and rapid-fire jargon can make them difficult to follow. Gemini workers used more than 20,000 hours of Watch Duty radio recordings to train the AI tool https://www.cnet.com/tech/services-and-software/how-to-stop-ai-chatbots-from-training-on-your-personal-data/ to recognize firefighting terms that aren’t in the average dictionary or books that AIs are typically trained on https://www.npr.org/2026/08/19/nx-s1-5936438/why-tech-companies-are-buying-up-tons-of-rare-old-books-to-train-their-ai-models . Google says the tool is not designed to replace human decision-making in life-or-death situations. “This tool takes on the demanding work of listening, freeing Watch Duty’s volunteers to focus on what matters most: confirming fires and getting warnings to people while they can still act,” said Stephanie Chew, Google project team lead. The project was completed at no cost thanks to a Google.org fellowship, and the resulting Watch Duty app features are free to use. Improving real-world fire alerts and transcriptions From the data collected, Gemini engineers and Watch Duty identified several measures of success. One metric was the word error rate, which fell by around 15% after the AI model was implemented. Another sign of success was a jump in Watch Duty volunteers’ satisfaction with their radio transcription tools, from around 35% to 92%. The Gemini tool https://www.cnet.com/tech/services-and-software/chatgpt-vs-gemini-which-chatbot-is-best-at-seeing-the-world/ continues to analyze approximately 200 hours of radio speech traffic from 400 feeds per day. The tool was made to handle more than 10,000 radio feeds if necessary. The integration could pave the way for broader use across public safety. The platform’s built-in scalability could eventually bring similar audio-clearing tools to local fire departments, 911 call centers and Coast Guard operations, where clear communication is critical under pressure.