Waymo Robotaxis Are Better Than Human Drivers. But Not Everywhere

• A new study from Waymo concludes that riding in a Waymo robotaxi in urban environments is safer than human drivers.
• Waymos had fewer crashes in intersections and fewer airbag deployments. Cyclists and pedestrians were also safer around these robotaxis.
• The study, however, has some drawbacks. 

Robots have already creeped into our lives in ways that would have sounded like science fiction a few years ago. From vacuuming our floors to AI companions with unholy use cases, machines with programmable eyes, ears and brains are here to stay, everywhere and in every form. But perhaps the biggest change is that multiple cities now have driverless taxis operating on public streets.

Now, we’re starting to see larger datasets indicating that self-driving cars—in many cases—may have better safety records than humans.

A new study from Waymo found that Alphabet’s Waymo driverless ride-hailing service is safer than human drivers in many real-world conditions. While the study was authored by Waymo engineers, it’s set to be published in the journal Traffic Injury Prevention, which means it’s been peer-reviewed by the scientific community for its methods.

Combing through over 56.7 million miles of driving data through January 2025, they found significantly lower crash rates than human drivers. Of that, 31.1 million miles were driven in Phoenix, 18.2 million miles in San Francisco, 6.4 million miles in Los Angeles and about 800,000 miles in Austin. 

In the mostly-sunny cities where Waymo currently operates, the study found that the largest improvement emerged from vehicle-to-vehicle crashes in intersections, where Waymo vehicles demonstrated a 96% reduction in any type of injuries reported. Driving performance was compared against publicly available data for drivers in Texas, California and Arizona.

The second largest improvement was in the overall number of airbag deployments, with Waymos 91% less likely to have an airbag event than a human driver. Note here two important caveats: In certain cities Waymos do not operate on the highway, reducing impact speeds, and since they don’t have a driver are less likely to have an occupant in a seat with a front air bag.  Injuries related to other crashes were also down, including those with cyclists (down 82%), motorcycles (down 82%) and pedestrians (down 92%).

Of course, humans still dominate the roads—and crash a lot. The U.S. Department of Transportation notes that over 42,000 lives were lost in 2022 on America’s streets and highways. But Federal Highway Administration data indicates that humans in aggregate drive nearly 100 million miles before a fatal accident occurs.

Waymo’s cars haven’t hit that benchmark yet, but over half that distance, they’re already proving to be safer. But they need to clock tens of millions of additional miles to show any statistically significant results on serious injuries and deaths.

So far, the dataset only covers mileage in cities where weather patterns don’t swing dramatically. Its debut in Washington D.C. next year will be worth keeping an eye on. It will be the company’s first deployment in the Northeast, where weather swings are wild with frequent snowstorms and fewer sunny days than California and Arizona. Still, D.C. is on the mild end of those weather patterns, but operating in Buffalo, New York should really test Waymo’s tech.

That said, Waymo has been more transparent about its safety data than Tesla, which is also planning to launch an unsupervised ride-hailing service in Austin next month. Tesla has made huge claims about the safety record of its Full Self-Driving (FSD) software, but there’s little publicly available data to corroborate those claims. And AI scientists have doubts.

That said, driverless taxis are coming to the cities near you, and even Uber and Lyft are on board with the concept.

Have a tip? Contact the author: suvrat.kothari@insideevs.com