AI Just Helped American Airlines Reduce Its Climate Impact

The solution wasn't a new fuel. It was a slightly different altitude.

American Airlines and Google have published results from a contrail-avoidance trial involving 2,400 flights between the US and Europe, in which AI-generated forecasts were used to predict where contrails were likely to form and to offer pilots optional routes to avoid those zones. For the 112 flights that took the alternative routing, contrail formation dropped 62% compared to the control group. The estimated reduction in climatological warming from those flights: 69%.

No new aircraft. No sustainable aviation fuel. No significant additional fuel burn.

Why Contrails Matter More Than They Look

Those thin white lines behind aircraft are responsible for 1% to 2% of total planetary warming, according to Contrails.org, the nonprofit research organization affiliated with Bill Gates' Breakthrough Energy group — which was among the collaborators on this trial. That figure is disproportionate to the seriousness with which contrails are discussed in aviation climate conversations, which tend to focus on carbon emissions and fuel type.

The mechanism is straightforward: when aircraft pass through cold, humid air, ice crystals form around engine soot particles, creating persistent cloud formations that trap heat. Some contrails dissipate quickly. Others persist for hours or days. Research suggests that minor adjustments to flight altitude or routing — enough to avoid the specific atmospheric conditions in which persistent contrails form — could eliminate a meaningful share of this warming effect at minimal additional cost.

The American Airlines and Google trial is the largest test of that theory conducted in the United States, according to Thomas Walker, senior transportation technology manager at the Clean Air Task Force.

How the AI Component Works

Google developed an AI-based forecasting tool that predicts where contrail-forming conditions exist along a given flight path. Those forecasts were integrated into Flightkeys, the flight planning software used by American Airlines, making contrail risk visible to dispatchers alongside existing route and fuel data.

The trial ran from January to May 2025. Half of the 2,400 flights were offered a contrail-avoidance route option; the other half served as the control group. Participation was optional — pilots and dispatchers could choose whether to file and fly the alternative plan.

Of the flights offered the option, 112 took it. That's a relatively small subset of the total trial group, and it points to one of the real-world constraints the research surface: adoption. The logistics of coordinating altitude adjustments across different airspace regions and international air traffic control systems adds friction, and some airlines have shown reluctance to participate in similar European trials.

American Airlines vice president of sustainability Jill Blickstein noted that the trial demonstrated it wasn't operationally difficult for dispatchers and pilots to file and execute the alternative plans. The deterrent most often cited by airlines — the cost of extra fuel on a longer or higher route — was not supported by the data. The trial found no statistically significant difference in fuel usage between the two groups.

What Comes Next

American Airlines has not yet incorporated contrail avoidance into its standard flight planning process. The airline has indicated it intends to continue working with Google, Flightkeys, and Contrails.org on additional studies covering different routes and times of day.

Google's contrails lead, Dinesh Sanekommu, described the goal as building contrail forecasting and avoidance features into all major flight-planning software providers — making what is currently a trial capability a standard layer of aviation operations. The North Atlantic corridor, where the trial was conducted, is a particularly high-impact zone for contrail formation, making it an effective testing environment.

The broader significance, as Walker at CATF frames it, is that this trial provides evidence for other airlines currently weighing the operational trade-offs. A large-scale, peer-reviewed demonstration that the fuel cost concern doesn't materialize in practice removes one of the primary arguments against adoption.

A Use Case Worth Attention

The contrail trial is notable not just for its climate implications but for what it illustrates about where AI generates practical value in complex physical systems. The tool doesn't redesign aircraft or change fuel chemistry. It processes atmospheric data at scale to surface information about where ice crystal formation is likely, making it actionable within an existing operational workflow.

That model: AI as a forecasting and decision-support layer integrated into systems operators already use, with measurable outcomes and low switching costs, is the implementation pattern that tends to produce durable results. It's less dramatic than the frontier model announcements that dominate AI coverage, and considerably more deployable.

For organizations thinking about where AI creates genuine operational value rather than speculative future returns, this kind of applied use case is worth studying. If you want help identifying where that pattern applies in your own business, Winsome Marketing's team is a good place to start.