Health · Public Health & Disease · 2 days ago
Google AI model best predicted US flu hospitalizations in 2025-26 season
The US CDC found that a Google AI model was the best of 39 at forecasting flu hospitalizations in the 2025-26 season.
It estimates hospitalizations for the current week and the next three weeks.
These forecasts can help states prepare medical services for busy periods.
Researchers can inspect how the model reaches its results, but people are still needed to guide the work.
The US Centers for Disease Control and Prevention said Google’s Empirical Research Assistance model performed best among 39 artificial intelligence models at forecasting flu hospitalizations during the 2025–2026 season.
The CDC report estimated hospital admissions for the current week and the following three weeks using data from government, industry and academic teams.
The forecasts can help predict demand for medical services in each state and support planning.
Google’s ERA system uses large language models to generate and test disease-forecasting models with limited direct human intervention.
A study published in Nature in May reported that ERA-generated models performed as well as or better than models used by the CDC for influenza, Covid-19 and respiratory syncytial virus.
Researchers said human teams remain important for guiding the hypotheses explored by the AI.
- Who
- The US Centers for Disease Control and Prevention and Google
- What
- Google’s ERA model was rated the best of 39 AI models at forecasting flu hospitalizations.
- When
- The 2025–2026 flu season; the report was published on October 9, 2026.
- Where
- The United States
- Why
- The forecasts can help plan for demand for medical services.
This story does not have two clearly opposing sides.
Esse desempenho valida nossa confiança de que a IA, combinada com a engenhosidade humana, aprimorará nossa capacidade de prever doenças em todo o mundo
This story does not have a timeline yet.
- Models assessed
- 39 AI models designed for healthcare
- Forecast period
- The current week and the next three weeks
- Google model
- Empirical Research Assistance (ERA)
- Diseases covered in the study
- Influenza, Covid-19 and respiratory syncytial virus (RSV)
- Study publication
- May, in Nature
Sources
Related news
Tokat suspends classes as thousands seek care for stomach symptoms
Singapore mosques to stay open for prayers as haze worsens
Survey finds rising heat-related risks for pregnant women and newborns
WHO says cause of Russian lab worker’s death remains unknown
Russia ends checks after Siberian lab worker’s death, but cause remains unknown
Chickenpox was the most common infection in South Korea’s military from 2022 to 2026
Community-led initiative to address drug abuse launched in 300 Jind villages
About 1,300 to 1,400 people seek care after illness in Tokat
Brazil approves first human tests of vaccine against a major malaria parasite
Drug-resistant Shigella is becoming more common in the U.S
Kenya’s First Ebola Patient Travelled Through DR Congo and Uganda Undetected
