Robotaxis Are Here. But Are They Really Safer Than Human Drivers?

Imagine ordering a ride, watching a car pull up to the curb, and opening the door to discover something strange: there’s nobody in the driver’s seat.

No steering wheel adjustments. No driver checking the mirrors. No awkward conversation about traffic.

Just a car that knows where you’re going and plans to get you there on its own.

What once seemed like science fiction is now happening on American streets. Companies like Waymo are operating fully driverless taxis in several major cities, and their services have expanded rapidly.

But as more people climb into cars controlled by computers, one question matters more than almost anything else: Are robotaxis actually safer than human drivers?

The research is starting to reveal some surprising answers.

Free weekly newsletter

Discover something fascinating every Tuesday.

The most interesting science isn't always the science that makes the headlines.
Every week, we dig into fascinating new research, strange scientific questions, surprising discoveries, and the evidence behind them.

No spam. 1-click opt-out. Privacy-first.

Millions of Miles Without a Human Behind the Wheel

For decades, engineers dreamed of building cars that could drive themselves. The challenge was enormous.

Human drivers make thousands of tiny decisions during every trip. We watch traffic lights, predict what other drivers might do, notice pedestrians, and react when something unexpected happens.

Teaching a computer to handle all those situations requires a combination of sensors, artificial intelligence, detailed maps, and sophisticated software.

Waymo’s robotaxis use cameras, radar, and lidar, a technology that uses laser light to measure distances and create a detailed picture of the surrounding environment.

Together, these systems help the vehicle detect cars, cyclists, pedestrians, and obstacles.

The car’s software then predicts how nearby objects might move and decides when to accelerate, brake, turn, or stop.

Unlike a human driver, the system doesn’t get sleepy, check its phone, or become distracted by a conversation.

But removing human mistakes is only useful if the computer can reliably handle the complicated world of real traffic.

That’s where the science gets interesting.

What Happens When Scientists Compare Robotaxis With Human Drivers?

One of the most important questions in autonomous vehicle research is surprisingly simple: How often do these cars crash compared with ordinary drivers?

A 2024 study published in Traffic Injury Prevention examined 7.1 million miles of fully driverless Waymo operations.

Researchers compared crash rates with estimates for human-driven vehicles.

The results were encouraging. Waymo vehicles had lower rates of certain crashes, including crashes involving reported injuries and those reported to police.

Then researchers examined a much larger collection of driving data.

A 2025 follow-up study expanded the analysis to 56.7 million miles of driverless operation. It again found lower crash rates for several outcomes when compared with human-driving benchmarks.

That is a significant development.

Self-driving cars have spent years being tested in simulations and controlled environments. Now researchers can study millions of miles of actual trips on public roads.

And at least for some types of crashes, the numbers suggest that automated driving can outperform human drivers.

Why Computers Might Be Better at Driving Than People

Think about the last time you drove somewhere.

Maybe you changed the radio station, glanced at your phone’s navigation screen, or felt tired after a long day.

Even careful drivers experience distractions.

Computers don’t have those problems.

A robotaxi can continuously monitor its surroundings through multiple sensors. Its software can process information from several directions at once without becoming bored or losing concentration.

That creates some important potential safety advantages.

For example, a self-driving system can be designed to maintain safe following distances, obey speed limits, and respond consistently to traffic signals.

It also doesn’t experience road rage, drive drunk, or deliberately take risks to save a few minutes.

Human behavior contributes to many serious traffic crashes, so reducing those behaviors could make roads safer.

However, computers introduce a different category of problems.

They must correctly interpret complicated environments, including unusual road layouts, unexpected obstacles, and situations that were difficult to anticipate during development.

A robotaxi may never become impatient, but it can still misunderstand what is happening around it.

The Strange Situations That Challenge Self-Driving Cars

City streets are full of surprises.

A pedestrian steps into the road unexpectedly. A construction worker directs traffic around a closed lane. Heavy rain makes road markings difficult to see.

Humans often handle these situations using experience, judgment, and social cues.

For an autonomous vehicle, each situation becomes a complicated problem involving perception, prediction, and decision-making.

Research published in Nature Communications in 2024 examined differences between autonomous and human-driven vehicle crash patterns.

The study highlighted how driving conditions can affect comparative safety, with autonomous vehicles showing different strengths and weaknesses depending on the circumstances.

This matters because a car that performs well on a sunny afternoon may face very different challenges at dawn, during bad weather, or on unfamiliar roads.

And not every self-driving system uses the same technology.

A Waymo robotaxi, a vehicle from another autonomous driving company, and a consumer car with driver-assistance features may operate very differently.

Their safety records must be evaluated separately.

When a Robotaxi Makes a Mistake, Who Is Responsible?

Autonomous vehicles also introduce a question that ordinary cars rarely raise: What happens when the driver is software?

In a traditional crash, investigators examine the actions of human drivers.

With a robotaxi, they may need to examine sensor recordings, software behavior, vehicle design, and decisions made by the company operating the system.

This has already become a major issue for regulators.

The National Highway Traffic Safety Administration (NHTSA) requires reporting of certain crashes involving automated driving systems.

These reports help federal officials identify possible safety problems and investigate incidents.

Some high-profile failures have demonstrated why oversight matters.

Cruise, another autonomous vehicle company, faced intense scrutiny following a 2023 incident in which a pedestrian was dragged by one of its driverless vehicles.

Zoox has also faced federal safety reporting and recall activity involving its autonomous vehicle technology.

These incidents show that the transition to driverless transportation involves more than developing software that can navigate traffic.

It also requires systems for detecting failures, investigating accidents, correcting problems, and protecting the public.

Could Robotaxis Eventually Save Thousands of Lives?

The potential public health implications are enormous.

Traffic crashes are a major cause of injury and death worldwide. In the United States alone, tens of thousands of people die on roads each year.

Even a modest reduction in serious crashes could prevent substantial suffering.

Imagine a future in which fewer people are injured by distracted driving, impaired driving, or dangerous speeding.

Autonomous vehicles could contribute to that future if their safety advantages continue as services expand.

But the benefits would depend on more than the technology itself.

Cities would need to consider how robotaxis interact with pedestrians, cyclists, emergency vehicles, and public transportation.

Researchers would also need to examine whether robotaxis change how much people travel by car.

For example, if inexpensive driverless rides encourage many more vehicle trips, the total number of miles driven could increase. That could affect congestion, pedestrian exposure, and overall transportation safety.

The public health question is therefore bigger than whether an individual robotaxi can complete a trip safely.

It is whether a transportation system with autonomous vehicles ultimately produces fewer injuries and deaths.

The Biggest Test Is Still Ahead

Waymo’s published safety results represent an important milestone, but expanding driverless transportation will create new challenges.

A robotaxi operating in a familiar service area has access to maps, road information, and extensive testing in that environment.

Expanding into new cities means encountering different weather, road designs, traffic patterns, and driving behaviors.

Snow-covered streets are different from dry pavement. Busy intersections are different from quiet suburban roads. A crowded school zone presents different challenges from an open highway.

Researchers will be watching whether safety improvements continue as autonomous vehicles operate across more of these environments.

The most valuable evidence will come from large amounts of real-world driving data, consistent crash reporting, and independent comparisons.

Would You Trust a Car Without a Driver?

The arrival of robotaxis marks a remarkable moment in transportation history.

For more than a century, cars have depended on human beings to watch the road, make decisions, and respond to danger.

Now machines are beginning to take over those responsibilities.

And the early results are more than a technological curiosity. Peer-reviewed research involving millions of miles of Waymo driving suggests that autonomous vehicles can achieve lower rates of certain crashes than human-driving benchmarks.

That doesn’t make robotaxis flawless. It means researchers can finally measure how well they perform in the real world.

The next time you see a car traveling down the street with nobody behind the wheel, consider what is actually happening.

A machine is observing the world, predicting human behavior, and making decisions that once required a person.

The biggest breakthrough may not be that computers have learned to drive. It may be that they are beginning to show they can avoid some of the mistakes humans make every day.