Are Robotaxis Safer Than Human Drivers? What the Data Says

For Waymo, the available evidence points to yes. Its driverless cars are involved in far fewer injury crashes per mile than human drivers in the same cities. Most of that evidence is produced by Waymo. The evidence cannot yet settle the question for fatal crashes, and it says nothing about other operators. For Tesla and Zoox, the federal crash database lists incidents but not the mileage needed to turn them into rates.

Is Waymo safer than human drivers?

Waymo's Safety Impact hub is the company's data page. It covers 271.3 million miles driven with no human driver through June 2026. Those miles were in Phoenix, the San Francisco Bay Area, Los Angeles, Austin and Atlanta. Against an average human driver covering the same distance in those cities, Waymo reports:

• 95% fewer crashes involving serious injury or worse
• 82% fewer crashes with an airbag deployment in any vehicle
• 82% fewer injury-causing crashes
• 93% fewer pedestrian injury crashes and 86% fewer cyclist injury crashes

A peer-reviewed study in Traffic Injury Prevention was published in May 2025. It examined 56.7 million miles through January 2025. It found statistically significant lower crash rates for injury, airbag and serious-injury outcomes. It found no significant increase in any of 11 crash types. Its disclosure statement says all six authors are Waymo employees.

In December 2024, Waymo and reinsurer Swiss Re reported 88% fewer property damage claims and 92% fewer bodily injury claims than a human baseline. The comparison covered 25.3 million miles. It used liability claims as a proxy for at-fault crashes. Waymo's side rested on nine property damage and two bodily injury claims. The company said the paper had been submitted to a journal.

What are the limits of Waymo's numbers?

Waymo says on its hub that no perfect like-for-like comparison between human and autonomous-vehicle data exists. Its crash counts come from reports it must file with federal regulators. The human benchmark relies on police reports. Waymo adjusts the human injury-crash benchmark for the estimated 32% of injury crashes that are never reported to police. It applies no adjustment for airbag or serious-injury crashes.

The comparison covers surface streets only. The cities involved see little snow, and time of day is not adjusted for. Waymo says it lacks the mileage to detect a statistically significant difference in fatal crashes alone. It does not report them separately. The benchmark is all human drivers in an area. Waymo says no public data allow a comparison with taxi drivers.

What do independent researchers and regulators say?

NPR reported in December 2025 on independent analyses by Ars Technica and the newsletter Understanding AI. It said they supported Waymo's claim to be safer than human drivers.

Broader academic work is more mixed. University of Central Florida researchers wrote in Nature Communications in June 2024. They compared 2,100 crash records involving automated or driver-assist vehicles with 35,113 involving human-driven ones. Their matched analysis found automated driving systems less likely to crash in most scenarios. But it put the odds 5.25 times higher at dawn or dusk and 1.98 times higher while turning. Those are odds ratios from crash records, not per-mile rates.

As of early October 2026, NHTSA had published operators' mandatory crash reports through August 17, 2026. The agency says the reports are not normalized by miles traveled or any other measure of exposure. It says they may be incomplete or unverified and call for caution when comparing companies.

Recalls and investigations show failures that crash rates miss. In December 2025, Waymo said it would file a voluntary software recall after its vehicles passed stopped school buses. Austin's school district had documented 19 such incidents, and Waymo said none caused injuries, NPR reported. In June 2026, Waymo recalled nearly 4,000 robotaxis after at least 13 incidents of vehicles entering highway sections closed for construction, TechCrunch reported. Waymo said no collisions or injuries resulted, according to TechCrunch.

On January 23, 2026, a Waymo vehicle struck a nine-year-old girl near a school in Santa Monica, California. That is according to a preliminary National Transportation Safety Board report cited by TechCrunch. The report said the vehicle had slowed from 17 mph to 6 mph first. She reported minor injuries, and NHTSA said a week later that it was investigating. In April 2026, a flash flood in San Antonio swept away a Waymo vehicle, and a recall followed in May.

What about Tesla and Zoox?

Data Tesla gave NHTSA showed 14 crashes involving its Austin robotaxis since their June 2025 launch, CBS News reported in February 2026. Two of them involved minor injuries, it reported. By September, TechCrunch wrote that Tesla had reported a few dozen incidents in Austin. Most were minor crashes with stationary objects, it wrote. Without mileage, those counts are not a rate.

Zoox began charging for rides in Las Vegas on August 10, 2026. NHTSA said the exemption is subject to enhanced oversight. TechCrunch notes the company has issued several recalls.

The bottom line

Waymo's record in its established cities is substantial public evidence that a driverless service can outperform average human driving on injury crashes. That evidence is company-led, tied to specific cities and surface streets, and silent on fatal crashes. It does not transfer to Tesla, Zoox or any other operator.