Planes don’t just produce fuel emissions. Under certain conditions, their engines leave condensation trails, known as contrails, which can persist in the atmosphere and contribute to warming. Now, Google and American Airlines are studying how to use artificial intelligence to avoid routes with the highest probability of producing them.
AI enters flight planning
In 2023, Google and American Airlines conducted an initial test with a small group of flights. Pilots received AI-generated forecasts to identify areas where contrails were more likely to form.
The result was a 54% reduction in the formation of these trails across 70 flights. It was an important demonstration, but there was still a problem: selecting the right flights and coordinating each operation required several hours of manual work.
What if this information appeared directly in the tools airlines already use to plan their routes? That is the focus of the new research.
A trial with 2,400 transatlantic flights
For this phase, Google’s contrail forecasts were integrated into American Airlines’ flight-planning software. This allowed the recommendation to avoid certain areas to become part of the airline’s usual operational process.
The trial included 2,400 transatlantic flights from the company’s standard schedule. Among the flights that successfully carried out the avoidance plans, the contrail formation rate fell by 62% compared with the control group.
The accuracy of this result matters. That 62% applies to the flights that actually followed the recommended plans, not necessarily to all 2,400 flights in the trial. Even so, the test shows that integrating AI into existing systems can be more useful than a standalone tool that requires additional coordination.
How does this type of forecast work?
An AI system can combine weather information, atmospheric data, and flight characteristics to estimate where the risk of contrail formation is highest. With that forecast, planners can assess small changes in altitude or route.
The goal is not to redesign every flight or eliminate trails under all conditions. The idea is to find operational alternatives that reduce their formation without significantly affecting safety, schedules, or fuel consumption.
The key is not only for AI to make a correct prediction. It must also turn that prediction into a recommendation an airline can apply quickly and at scale.
From an experimental test to an operational tool
The difference between the two studies reveals one of the common challenges of applying AI to real-world problems. A model may work in a controlled demonstration, but its impact grows when it is integrated into the workflows where people already make decisions.
In this case, automating flight identification and route planning could reduce manual effort. It would also make it possible to evaluate contrail avoidance as a regular practice, rather than treating it as a special project for a small number of flights.
Questions still remain. Researchers will need to study how the system performs across different seasons, routes, atmospheric conditions, and operational constraints. In addition, any trajectory change must keep aviation safety as the top priority and assess its potential effects on fuel consumption and direct emissions.
The result does not turn AI into a complete solution for aviation’s climate impact. But it does show something valuable: prediction technology can move from the laboratory into daily operations when it is connected to the tools and decisions that are already part of the industry.
If these tests continue to confirm the results, avoiding contrails could become one of the complementary measures for reducing the climate impact of flights—without waiting for a complete transformation of aircraft or fuels.