AI in Healthcare & Science · AI in Public Health

Can AI Predict the Spread of a Pandemic?

AI can help generate probabilistic forecasts about how a disease might spread based on patterns in available data, and these models have been used to inform public health planning, but predictions carry significant uncertainty, especially for novel pathogens with limited historical data, and are not guaranteed to accurately forecast how an actual pandemic will unfold.

Medical disclaimer

This page is for general educational purposes only and is not medical advice. It does not replace a consultation with a licensed physician, pharmacist, or other qualified health provider. Always talk to your own care team before starting, stopping, or changing any medication or supplement.

Key takeaways

AI and machine learning models are used alongside traditional epidemiological modeling techniques to help forecast aspects of disease spread, such as case trajectories or resource needs.
These models generate probabilistic estimates based on patterns in available data, not certain predictions of exactly how a pandemic will unfold.
Forecasting accuracy tends to be more limited for novel pathogens with little historical data, and can be affected by human behavior changes, policy interventions, and other unpredictable factors.
Public health agencies have used pandemic modeling, including AI-assisted approaches, as one input for planning purposes, alongside other forms of expertise and judgment.
Past pandemic modeling efforts, including during COVID-19, illustrated both the value and the significant uncertainty involved in this kind of forecasting.

Forecasting, Not Fortune-Telling

AI and machine learning tools have become part of the broader toolkit epidemiologists and public health researchers use to model and forecast aspects of how a disease might spread through a population — things like projected case counts, hospital resource needs, or the potential impact of different intervention strategies. These models work by analyzing patterns in available data, including how the disease has spread so far, characteristics of the pathogen, and factors like population density and mobility patterns, to generate probabilistic projections about likely future trajectories.

It’s important to frame this accurately: these are forecasts based on patterns and assumptions built into a model, generating a range of probable outcomes, not certain predictions of exactly how an actual pandemic will unfold. Real-world disease spread depends on an enormous number of factors, many of which are inherently difficult to predict with precision.

Why Uncertainty Is Baked Into These Models

Several factors limit how confidently any model, AI-assisted or otherwise, can predict pandemic spread. For a genuinely novel pathogen, there’s limited historical data specifically about that pathogen’s behavior, meaning models often have to rely on assumptions drawn from related diseases or general principles of disease transmission, which introduces additional uncertainty, particularly in a pandemic’s early stages before enough direct data has accumulated. Human behavior is another major source of unpredictability — how people respond to a spreading disease, whether through voluntary behavior changes or policy interventions like public health guidance, can significantly alter the actual trajectory of spread in ways that are difficult to fully anticipate in advance. Because of this, forecasts further into the future generally carry more uncertainty than very short-term projections, and models are often updated frequently as new data becomes available.

Lessons From Recent Experience

Pandemic modeling efforts during recent public health emergencies, including COVID-19, illustrated both the genuine value and the real limitations of this kind of forecasting. Models provided useful input for public health planning and resource allocation decisions, helping inform choices under significant uncertainty, but forecasts also sometimes diverged from how events actually unfolded, particularly over longer time horizons or in the face of unpredictable factors like shifts in behavior or the emergence of pathogen variants. This experience reinforced among public health experts the importance of treating model outputs as one input for decision-making, alongside direct surveillance data and epidemiological expertise, rather than as a definitive prediction to be relied upon exclusively.

Bottom Line

AI can help generate probabilistic forecasts about pandemic spread that inform public health planning, but these predictions carry significant inherent uncertainty, especially for novel pathogens or longer time horizons, and are best understood as one useful input among several rather than a reliable, certain prediction of how an actual pandemic will unfold.

Go deeper

Important caveats

Pandemic spread models are inherently uncertain and can be significantly affected by future human behavior, policy decisions, and other factors that are difficult to predict in advance.
Model predictions should generally be understood as one input for planning and decision-making, not a guaranteed forecast of future events.

Frequently asked questions

How accurate were AI and modeling predictions during the COVID-19 pandemic?

Modeling efforts during COVID-19, including some that incorporated AI and machine learning techniques, showed a range of accuracy depending on the specific model, the time horizon being forecast, and how quickly the situation was evolving. Forecasts further into the future, or made early when less was known about the pathogen, generally carried more uncertainty than shorter-term forecasts made with more established data.

Why is it harder to predict the spread of a completely new pathogen?

Predictive models generally rely on patterns from historical data, and a genuinely novel pathogen may behave in ways that don't closely match prior diseases used to train or calibrate a model, meaning there's less relevant historical information available to ground predictions, which tends to increase uncertainty especially in the early stages of a new outbreak.

Do public health agencies rely solely on AI models for pandemic planning?

No. Public health agencies generally use modeling, including AI-assisted approaches, as one input among several for planning purposes, combined with epidemiological expertise, direct surveillance data, and broader public health judgment, rather than relying on any single model or method exclusively.

Sources

[1]Centers for Disease Control and Prevention — Centers for Disease Control and Prevention
[2]World Health Organization — World Health Organization

Written by Editorial Team

Last updated July 25, 2026

Get one well-sourced answer a week

No spam. Unsubscribe anytime.