how many autonomous cars are on the road

[McDelaney]

The number of autonomous vehicles (AVs) on public roads in the United States is relatively small and consists almost entirely of test vehicles, not consumer-owned cars. As of late 2023, estimates suggest there are fewer than 2,000 vehicles equipped with Level 4 automation technology actively being tested. The truly meaningful figure isn't the total count but the number of miles these vehicles drive autonomously, which for industry leaders like Waymo and Cruise, has reached millions of miles annually.

A key point of confusion is the difference between advanced driver-assistance systems (ADAS), like Tesla's Autopilot, and true autonomy. Systems requiring driver supervision are classified as Level 2 and are found in millions of cars. Fully autonomous vehicles, which operate without a human driver (Level 4/5), are the domain of specialized fleets from companies like Waymo, Cruise, and Zoox.

Here’s a breakdown of vehicles with higher levels of automation based on publicly reported fleet sizes and testing permits:

Company / Program	Estimated Fleet Size (Late 2023)	Automation Level	Primary Location(s)	Key Data Point (Miles Driven)
Waymo (Google)	700+ vehicles	Level 4 (Fully Autonomous in Geofenced areas)	Phoenix, San Francisco, Los Angeles	Over 1 million autonomous miles per year
Cruise (GM)	400+ vehicles (pre-incident)	Level 4 (Fully Autonomous in Geofenced areas)	San Francisco (operations paused)	Logged over 1 million driverless miles
Zoox (Amazon)	100+ vehicles	Level 4 (Testing & employee shuttle)	San Francisco, Las Vegas, Seattle	Testing in multiple cities with safety drivers
Nuro	150+ delivery pods	Level 4 (Goods-only delivery vehicles)	Houston, Phoenix, Mountain View	Commercial delivery partnerships
Various Automakers (e.g., Ford, Mercedes)	500+ vehicles (combined estimate)	Level 2+/Level 3 (Driver Assist / Conditional Automation)	Nationwide testing	Data collected for next-generation systems

The growth is constrained by significant technological hurdles, high costs, regulatory approval, and, crucially, public and governmental trust. The timeline for a significant number of consumer-owned autonomous cars on the road remains uncertain, likely more than a decade away.

[OJonathan]

Honestly, most "self-driving" cars you hear about are just test vehicles. I follow the tech closely, and the real number of driverless cars—ones without a safety driver—is probably in the low hundreds, mostly in a few specific cities like San Francisco and Phoenix. The rest are just fancy driver-assist systems that still need you to pay attention. The real count is tiny, but the miles they're logging is the impressive part.

[OSamantha]

From a safety and regulation standpoint, the number is intentionally limited. As someone who reads NHTSA reports, these vehicles operate under strict testing permits. There isn't a national registry, but based on state-level disclosures from California, Arizona, and Texas, the total fleet of permitted Level 4 test vehicles is well under 2,000 units. The focus for regulators isn't on volume but on verifying the safety and performance data from each million miles driven before considering wider deployment.

[SanElliot]

I live in Austin, and you'll see them occasionally—those weird-looking cars covered in sensors. But it's not like there's one on every corner. It feels more like a pilot program than a real transportation change. I'd guess there are maybe a few dozen here. The companies are being super cautious, especially after a few high-profile incidents. So, the number on the road is small and growing very, very slowly.

[LaDawson]

The answer depends entirely on your definition. If you count every car with adaptive cruise control and lane-keeping, that's millions. But true "hands-off, eyes-off" autonomy? That's a different story. The industry uses SAE Levels. Most so-called autonomous cars are Level 2, meaning the driver is still responsible. The jump to Level 4 is massive, and those vehicles number only in the hundreds, confined to specific mapped areas. The technology is here, but the scalable, widespread deployment is not.