The top model and the CDC ensemble are different results
Google Research says its AI-assisted flu forecast was best at predicting U.S. flu hospital admissions during 2025–26. The CDC evaluation adds a distinction: Google_SAI-FluEns was the top individual submission, while the combined FluSight ensemble used in CDC messaging ranked seventh among 39 eligible models.
Both statements can be true. Google’s model led individual submissions; the CDC’s blend had its own rank. The score is a season-wide comparison, not a claim that a model predicted every state and week perfectly. Sources: https://www.cdc.gov/flu-forecasting/evaluation/2025-2026-report.html and https://blog.google/innovation-and-ai/models-and-research/google-research/google-science-ai-flu-forecasts/
What FluSight was trying to predict
FluSight collects weekly forecasts of hospital admissions for the current week and up to three weeks ahead. The CDC uses the results to communicate likely demand for health services.
Thirty-nine models met the CDC’s participation criteria; 33 beat a baseline that carries forward the previous week’s admissions. Google says its forecasts were developed using Empirical Research Assistance, an AI tool for generating optimization algorithms in scientific research.
A useful result, not a personal health forecast
Hospital admission forecasting gives public-health teams a signal about demand across places and weeks. It does not predict whether a particular person will get sick. CDC notes that forecast performance fell around periods of rapidly changing trends.
The practical value is planning for staffing, beds and treatment resources. The seventh-place ensemble result also shows why “AI ranks first” needs context: the top individual model and the blended forecast are different systems.
Source published September 30, 2026. Coverage is based on the maker’s announcement and demonstration.