As a species, we’ve reshaped the world to suit our ambitions – building cities that scrape the skies and digital networks that pulse with life. But in doing so, we’ve also filled the air with invisible threats. The atmosphere of opportunity surrounding us is often tainted – thick with pollutants that quietly infiltrate our lungs.

Urban air pollution is a defining challenge of our time – complex in source, dynamic in behaviour and increasingly beyond the reach of traditional tools.

But as cities grow smarter, a powerful new ally is emerging in the fight for cleaner air: artificial intelligence (AI).

What if AI could help us see the invisible? What if it could help us breathe more easily, live longer and reduce the health toll of invisible, toxic exposures?

AI offers a radically new toolkit to tackle this problem – analyzing vast datasets, uncovering hidden patterns, predicting future conditions and optimizing how our cities function. This article explores how we can use it to monitor, manage and ultimately mitigate urban air pollution concentrations and exposure in our rapidly urbanizing world.

Billions of people worldwide breathe air that exceeds World Health Organization (WHO) guidelines for pollutants like particulate matter (PM2.5) and nitrogen dioxide (NO2). This results in over 8 million premature deaths annually and costs the global economy $8 trillion each year.

Historically, our understanding of air pollution has relied on a relatively sparse network of fixed, costly monitoring stations. While reference-grade monitors and supersites (intensive monitoring stations with advanced capabilities) provide valuable data – such as chemical composition and particle size – they offer limited insight into hyper-local exposure – the variations people experience as they move through cities.

Today, that’s changing, thanks to AI, modelling and sensor networks.

Early applications of machine learning in air quality focused on models such as random forest, gradient boosting and hybrid approaches. These enable researchers to better predict concentrations of air pollution in cities while disentangling the effects of confounding variables like weather.

This predictive capability unlocks new insights into the effectiveness of policy interventions.

More recently, machine learning approaches have been used to improve the quality of air quality sensor measurements and apportion complex pollution sources, analyzing datasets that integrate pollution levels, weather conditions, traffic flows, industrial operations and more.

Combined with sensors, satellite imagery and even crowdsourced data, AI can now generate high-resolution, near-real-time pollution maps to track exposure and inform public health action.

One example is the DyNA system developed at Imperial College London, which combines physical modelling with AI techniques. It uses a customized Recurrent Neural Network to process time-series data and forecast pollution events.

When coupled with data assimilation techniques, DyNA can ingest real-world observations to enhance the accuracy and speed of air quality predictions. Google’s Air Quality API also combines diverse inputs to estimate global pollutant concentrations.

Smarter systems mean smarter decisions – not just for cities but for citizens too.

Imagine a runner or a cyclist receiving real-time alerts on their smartwatch, identifying pollution hotspots along their route and being advised to slow down or switch sides of the street in a narrow canyon where pollutants accumulate. These once-invisible insights are now possible.

AirTrack, developed by Air Aware Labs, has helped athletes and commuters reroute based on real-time exposure data. It combines GPS data, AI, air pollution modelling and user behaviour to deliver personalized insights into air pollution exposure, empowering people to make smarter choices.

For example, rerouting a cycle commute, adjusting the timing of a jog or simply deciding the best time to go outside. With remarkable precision, it will soon be possible to estimate where every particle you breathe has been emitted, opening up new frontiers in accountability, health monitoring and prevention.

Over time, AI will guide entire urban systems – transport, energy, land use – towards smarter, cleaner equilibrium. That includes dynamic road pricing, mobility-on-demand, and electric autonomous vehicles that are aligned with air quality goals.

By modelling how buildings affect airflow, emissions and pedestrian movement, AI tools help avoid the creation of pollution hotspots and prioritize health in design – from the first blueprint to the final build. This represents a shift from “reactive” to “proactive” planning, enabling healthier cities by design.

The potential of AI is immense but so are the risks if it’s deployed irresponsibly. Air pollution hits the poorest hardest and AI could widen these inequalities. For instance, due to missing infrastructure, Google’s API lacks coverage in many African cities.

For AI to be a force for good, it must be:

AI must not be done to people but built with them. That means opening up access to data, supporting open-source development and involving local voices in how tools are designed and deployed.

At the same time, AI is an essential enabler if we are to meet global targets such as the World Health Organization’s goal to reduce the health impacts of air pollution by 50% by 2040, set out at the Global Conference on Air Pollution and Health.

AI offers transformative power in the fight against urban air pollution. It enhances how we measure, predict, optimize and act, giving us the tools to make cities smarter, cleaner, fairer and more liveable.

These advancements could align with initiatives such as the World Economic Forum Alliance for Clean Air and C40’s commitment to air quality co-benefits through climate action.

But realizing this potential won’t come from algorithms alone. It requires intentional collaboration between governments, researchers, startups, city leaders and everyone breathing the air each day.

Therefore, philanthropists, multilateral institutions and the private sector must work together to close data gaps in underrepresented regions because clean air should not depend on postcodes or gross domestic product.

Our air should not be a hidden hazard. With AI as a partner and equity as our compass, we can build smart cities where clean air flows as freely as the data that drives them.

https://www.weforum.org/stories/2025/04/the-role-of-ai-in-reducing-urban-pollution/