Driverless cars offer genuine promise, but whether they’re a “good idea” depends on which problem you’re looking at. They could prevent thousands of deaths caused by human error, open up transportation for millions of people with disabilities, and eventually smooth out traffic. They also introduce new risks: empty vehicles clogging roads, massive job losses in trucking, unresolved legal questions about who pays when things go wrong, and technology that still can’t handle rain or snow reliably. The honest answer is that driverless cars are a good idea that isn’t fully ready yet, and the transition period will be messy.
The Safety Promise Is Real but Complicated
The strongest argument for driverless cars starts with a single statistic: drivers are the critical cause of 94% of all crashes, according to NHTSA’s national crash causation survey. That includes distracted driving, drunk driving, fatigue, and simple misjudgment. A computer doesn’t get tired, check its phone, or drive home from a bar. In theory, removing the human from behind the wheel could prevent the vast majority of the roughly 40,000 annual traffic deaths in the United States.
In practice, the safety picture is murkier. A study published in Accident Analysis & Prevention found that autonomous vehicles operating in self-driving mode were rear-ended at 4.8 times the rate of conventional cars. In urban environments specifically, that figure rose to 5.0 times. The likely explanation is that self-driving cars follow traffic rules precisely, braking and yielding in situations where human drivers expect others to be more aggressive. That cautious behavior, while technically correct, creates surprises for the human drivers sharing the road. This gap may shrink as autonomous vehicles become more common and other drivers adjust, but for now, mixing self-driving and human-driven cars on the same roads introduces friction.
Traffic Could Improve, but Not Right Away
Self-driving cars that communicate with each other can, in theory, drive closer together at more consistent speeds, reducing the stop-and-go waves that cause congestion. A simulation study modeling traffic in Qingdao, China, found that when autonomous vehicles made up more than 20% of traffic, overall efficiency started to improve. Once they reached 50% of traffic, their ability to coordinate in groups pushed urban traffic flow meaningfully higher. At 90% penetration, average speeds climbed to about 28 mph compared to 25 mph for manually driven vehicles on the same roads.
The catch is the transition. We’re nowhere near 20% autonomous traffic on public roads, let alone 50%. During the long period when self-driving and human-driven cars share lanes, the benefits are minimal. And there’s a counterintuitive problem: driverless cars could actually increase the total number of miles driven. Simulations consistently show that for every 100 miles a self-driving fleet travels, 15 to 25 of those miles are “empty,” with no passengers at all. These are repositioning trips, vehicles driving themselves to pick up the next rider or heading to a charging station. A study of Bloomington, Illinois estimated 17% of fleet miles were empty. In Japan, researchers projected that autonomous vehicle adoption could increase total miles driven by 22% to 44% by 2040, with a corresponding rise in CO2 emissions.
Without regulations to limit empty cruising, driverless cars could make congestion and emissions worse, not better, especially in dense cities.
Millions of People Could Gain Independence
One of the most compelling and least-discussed benefits is what driverless cars mean for people who can’t drive. The Ruderman Family Foundation estimated that removing transportation barriers for people with disabilities would open employment opportunities for roughly 2 million individuals and save $19 billion annually in healthcare costs from missed medical appointments alone. For elderly people who’ve lost the ability to drive safely, autonomous vehicles could mean the difference between isolation and staying connected to their community, medical care, and daily errands.
This is the kind of benefit that’s hard to capture in a cost-benefit spreadsheet but represents a real quality-of-life transformation for tens of millions of people.
The Job Displacement Is Enormous
About 3.5 million people work as truck drivers in the United States, and long-haul trucking is one of the first sectors autonomous technology is targeting. The routes are simpler: long stretches of interstate highway with predictable conditions. Projections estimate that by 2030, 500,000 to 875,000 long-haul drivers on major corridors like I-10, I-40, and I-80 could lose traditional roles. By 2040, autonomous trucks may handle 65% to 75% of freight that previously required human drivers, eliminating 2.3 to 2.6 million trucking jobs.
These aren’t entry-level positions. For many drivers, trucking is a middle-class career that doesn’t require a college degree. The economic ripple effects extend to truck stops, roadside motels, and rural communities built around freight corridors. New jobs will emerge in fleet management, vehicle maintenance, and remote monitoring, but they’ll require different skills and likely be located in different places.
The Technology Still Has Hard Limits
Current self-driving systems rely on a combination of cameras, laser-based sensors (lidar), and radar. Each has specific weaknesses. Cameras can’t recognize lane markings or traffic signs when they’re covered in snow. Lidar sensors malfunction when precipitation is falling. Researchers at MIT have been working on ground-penetrating radar that reads the road’s subsurface as a workaround, but this technology isn’t commercially deployed.
This means driverless cars perform well in the sunny, well-mapped streets of Phoenix or San Francisco but struggle in conditions that are routine for much of the country: heavy rain, snow, construction zones, and unpaved roads. Between 2015 and 2022, across 12 autonomous vehicle manufacturers testing in California, there were over 100,000 instances where a human had to take control from the automated system across roughly 15 million kilometers of testing. The rate of these interventions has improved over time, but it underscores that the technology is still being refined.
Nobody Agrees on Who’s Liable
When a fully driverless car causes a crash, the legal question of fault gets genuinely difficult. There may be no human driver to blame. The software did what it was designed to do, and the crash happened anyway. Current product liability law requires proving a defect, but what if the system simply wasn’t capable of handling a situation that no existing software could handle?
Legal scholars have proposed several frameworks. One approach would hold manufacturers strictly liable for any damage their cars cause, regardless of defect, ensuring victims are always compensated. Another would compare the car’s behavior to what a competent human driver would have done in the same situation. A third would compare it to the performance of other autonomous systems on the market, essentially asking whether the car met the industry standard. Each framework creates different incentives for manufacturers and different outcomes for crash victims. No unified federal standard exists yet, leaving a patchwork of state-level rules.
The Public Isn’t Convinced Yet
Even as the technology advances, trust remains a major barrier. A 2025 AAA survey found that 6 in 10 U.S. drivers are still afraid to ride in a self-driving vehicle. While 74% of drivers were aware that robotaxis exist, 53% said they wouldn’t choose to ride in one. Fear of the technology tends to decrease with exposure (people who’ve actually ridden in autonomous vehicles report less anxiety), but widespread adoption requires a level of public confidence that hasn’t materialized.
This skepticism isn’t irrational. The technology is improving rapidly but hasn’t yet demonstrated the kind of overwhelming safety record that would make handing over control feel like an obvious choice. For driverless cars to become a genuinely good idea in practice rather than just in theory, the gap between their potential and their current performance, along with the legal, economic, and environmental questions they raise, needs to close considerably.