Google says integrated AI contrail planning cut formation rates 62% on 2,400 flights

Google said on Mar 19, 2026 that its contrail-avoidance research has moved from a small pilot into regular airline operations. In 2023, Google and American Airlines showed that AI-based forecasts could reduce contrails by 54% across 70 test flights, but the process still required manual coordination to identify the right routes and flights.

The new result is larger and more operational. Google integrated its AI contrail forecasts directly into American Airlines’ existing flight-planning software and ran a trial covering 2,400 transatlantic flights that were part of the carrier’s standard schedule. For flights that actually flew the contrail-avoidance plans, Google says the contrail formation rate fell 62% compared with the control group.

Contrails matter because their warming effect can be significant even though they are not the same thing as direct carbon emissions. The technical problem is not only predicting where atmospheric conditions are likely to produce contrails, but also turning that forecast into a route-planning decision without adding too much operational friction. Google’s argument is that the bottleneck is as much workflow integration as model quality.

• 2023 pilot: 54% reduction across 70 flights
• 2026 trial: 2,400 scheduled transatlantic flights
• 62% lower contrail formation rate on flights that followed the avoidance plan
• Forecasts embedded into existing flight-planning tools instead of manual coordination

This matters because it suggests climate-oriented AI can move from demonstration projects into day-to-day operations if it is inserted into the software stack airlines already use. Google also describes contrail avoidance as scalable and cost-effective, which is important: environmental interventions in aviation typically fail when they add too much friction to dispatch or fuel planning.

The main caveat is that the 62% figure applies to flights that successfully executed the contrail-avoidance plan, not every flight in the broader schedule. Even so, the study is a strong proof point that AI forecasting plus workflow integration can reduce a real-world aviation externality without waiting for a new aircraft generation. For AI in climate and applied science, that is the bigger signal.