Waymo counts 94% fewer serious injury crashes than human drivers, on a comparison that leaves freeways out

The company figures

Waymo runs cars that drive themselves with no human in the driving seat, and it counts those trips as rider only miles. Through March 2026 it reports 220.6 million of them. Over that distance it says it had 94% fewer crashes in which somebody was seriously injured or killed than human drivers would have had, 47 fewer such crashes, and 82% fewer injury crashes, 707 fewer. The counts include every crash regardless of who caused it. A serious injury here means a police report recorded somebody as incapacitated or killed.

What the 94% is measured against

The comparison is not with all American drivers. Waymo builds a benchmark, meaning the crash rate it expects a human to have over the same distance, from state crash records in Maricopa County, San Francisco County, Los Angeles County and Travis, Hays and Williamson Counties in Texas. It counts passenger cars only, and only on surface streets, meaning ordinary roads rather than freeways. Freeway driving is excluded from both sides. The benchmark is then reweighted to the parts of those counties Waymo actually drives in, so an area it never enters is weighted out. Phoenix alone carries 80.6 million of the 220.6 million.

The human rate depends on crashes nobody reported

Human crash rates come from police reports, and many crashes never reach one. The Waymo study raised its human injury benchmark by 33% to cover injury crashes never reported to police, and applied no such correction at all to the airbag and serious injury comparisons. Noah Goodall, a researcher at the Virginia Transportation Research Council, reviewed 8 crash databases in the Journal of Safety Research and found estimates of missing crashes running from 15.4% to 84% depending on the method. A larger correction makes human drivers look worse and widens the Waymo lead. Goodall also notes that the federal reporting order the Waymo crash counts come from collects no mileage at all, so the miles in the denominator are supplied by Waymo.

The peer reviewed number rests on 2 crashes

The study covering the first 56.7 million miles to January 2025, accepted for publication in the journal Traffic Injury Prevention, counted 48 injury crashes, 18 in which an airbag went off, and 2 in which somebody was seriously injured. Every author is a Waymo employee, which the paper states. It puts the serious injury reduction at 85%, and its 95% confidence interval, the range of values the data is consistent with, runs from 39% to 99% fewer.

The results are statistically significant but because serious injury cases are, fortunately, rare, they're based on a small number of events.

Waymo, in its published statement about the study. Source 5.

Both serious crashes started with another driver. In one the Waymo car was stationary in traffic. Not every crash type improved. The same paper records a statistically significant rise in Phoenix in injury crashes where the Waymo car was struck from behind, 8 more than an average human over the study period.

A federal investigation is open

The Office of Defects Investigation at the National Highway Traffic Safety Administration opened a preliminary evaluation, PE26001, on January 28, 2026, covering an estimated 3,067 vehicles. A preliminary evaluation is the first stage of a federal defect inquiry and carries no finding. On January 23, 2026 a Waymo car with no safety operator on board struck a child 2 blocks from a Santa Monica elementary school during drop off hours, after the child ran into the street from behind a double parked vehicle. Waymo reported minor injuries.

It's encouraging to see real-world data showing Waymo outperforming human drivers when it comes to safety.

Jonathan Adkins, chief executive of the Governors Highway Safety Association, the body that represents state highway safety offices. Source 5.