A forecasting model that Google built with its AI was the most accurate predictor of flu-related hospital admissions in the 2025-26 season, according to an end-of-season analysis the Centers for Disease Control published this week. Of 39 eligible models, Google's submission came closest to the season's observed hospital admissions, the CDC found.
The forecasts were generated with Empirical Research Assistance, an AI tool that produces optimization algorithms for scientific work, Google says. Research on ERA was published recently in the journal "Nature," and the technology behind it is now available to trusted testers through Google's experimental science tools.
Read more: Google's WeatherNext 3 Forecasts Every Hour at Higher Resolution
The CDC's FluSight effort runs from October through May, gathering weekly submissions from government, industry, and academic teams that predict U.S. hospital admissions for the current week and three weeks ahead. Each week the agency combines those forecasts to communicate expected state-level demand for medical services.
Google framed the result as evidence that pairing AI with human expertise can sharpen disease forecasting worldwide. The announcement does not say how many teams took part beyond the 39 eligible models, nor whether the model will be submitted again for the coming season.
The notice also does not disclose pricing or a general availability date for ERA, which remains limited to trusted testers for now.
