San Francisco has become the world's most active testing ground for fully autonomous ride-hailing. If you've searched for driverless cars San Francisco, you've likely seen viral videos of steering wheels turning on their own or white minivans navigating the city's steep hills without a human behind the wheel. But what's the reality on the ground in 2026? Which services can you actually hail, how safe are they, and what does a robotaxi ride actually cost compared to Uber or Lyft?
This guide cuts through the hype. We'll walk you through the exact services operating in the city, the real safety data (not marketing claims), the technology making it possible, and the hard limitations you need to know before you book your first driverless ride.
Is This Truly Self-Driving? Understanding SAE Automation Levels
The single most important distinction in the autonomous vehicle space is the difference between consumer driver-assist systems and true robotaxis. Tesla's Full Self-Driving (Supervised), GM's Super Cruise, and Ford's BlueCruise are all SAE Level 2 systems — the driver must remain engaged, hands on or near the wheel, eyes on the road, and legally responsible for the vehicle at all times.
The robotaxis operating in San Francisco — primarily Waymo and, in limited capacity, Zoox — are SAE Level 4 vehicles. There is no human driver. No one is monitoring the road. The vehicle handles all driving tasks within its defined operational design domain (ODD), and if it encounters a situation it cannot handle, it is designed to pull over safely rather than hand control back to a passenger.
| Level | Name | Who Drives? | Examples |
|---|---|---|---|
| Level 2 | Partial Automation | Human (always) | Tesla Autopilot/FSD, GM Super Cruise, Ford BlueCruise |
| Level 3 | Conditional Automation | System (limited), human on standby | Mercedes DRIVE PILOT (highway only, ≤40 mph) |
| Level 4 | High Automation | System (within ODD), no human needed | Waymo, Zoox (geofenced urban areas) |
| Level 5 | Full Automation | System (everywhere, all conditions) | Does not exist commercially |
The takeaway: when you step into a Waymo in San Francisco, you are riding in a vehicle with no driver. That is a fundamentally different proposition — legally, technically, and experientially — from a Tesla on Autopilot with a human in the driver's seat.
Where Can You Use Driverless Cars in San Francisco Right Now?
As of early 2026, San Francisco has two primary robotaxi operators, though their service footprints differ significantly.
| Service | Vehicle | SF Coverage | Hours | Status |
|---|---|---|---|---|
| Waymo | Jaguar I-PACE (EV, custom 5th-gen sensor suite) | Most of SF proper, including downtown, SoMa, Mission, Sunset, Richmond; SFO airport rides available | 24/7 | Fully operational, open to public via Waymo One app |
| Zoox | Zoox purpose-built carriage (no steering wheel) | Limited geofenced zones in SoMa and downtown core | Limited hours | Invite-only / limited public access; expanding gradually |
Waymo is the dominant player. After receiving approval from the California Public Utilities Commission (CPUC) for 24/7 commercial operations in mid-2023, Waymo steadily expanded its San Francisco service area throughout 2024 and 2025. As of 2026, the service covers most of the city and includes trips to San Francisco International Airport (SFO). The Waymo One app works like Uber or Lyft — enter your destination, get a price, and a vehicle arrives.
Zoox, Amazon's autonomous vehicle subsidiary, has been testing its purpose-built carriage — a compact, bidirectional vehicle with no steering wheel or pedals — on San Francisco streets. However, public availability remains restricted compared to Waymo. Zoox operates under a different regulatory framework and has focused heavily on Las Vegas for its public-facing service.
Cruise (now defunct in SF): It's worth noting that Cruise, the GM-backed robotaxi operator that once ran hundreds of vehicles in San Francisco, suspended all operations in late 2023 after a serious incident in which a pedestrian was dragged under a Cruise vehicle. GM ultimately shut down Cruise's robotaxi operations entirely in 2024. Cruise no longer operates in any city.
How Much Does a Robotaxi Ride Cost vs. Uber or Lyft?
Pricing is one of the most common questions about driverless cars in San Francisco. Here's what you can expect in 2026:
| Route (approximate) | Waymo | UberX | Lyft Standard |
|---|---|---|---|
| SoMa to Mission (3 mi) | $12–$16 | $10–$18 | $10–$16 |
| Downtown to Sunset (7 mi) | $20–$28 | $18–$30 | $16–$28 |
| SoMa to SFO Airport (14 mi) | $38–$50 | $35–$55 | $32–$50 |
The bottom line: Waymo pricing is broadly competitive with Uber and Lyft for standard rides. It is not dramatically cheaper — Waymo is not yet subsidized to undercut rideshare on a per-ride basis in the way some early projections suggested. However, Waymo pricing is more consistent: surge pricing, which can double or triple Uber/Lyft fares during peak demand, does not apply to Waymo in the same way. You'll see a quoted price and it generally holds.
Waymo does not currently offer a shared/pool option comparable to Uber Pool, so for solo riders on short trips, a standard UberX or Lyft might occasionally be cheaper. For airport runs and medium-distance trips, Waymo is often within a few dollars of its competitors — and you get the novelty (and, for some, the comfort) of no human driver.
How Safe Are Robotaxis? The Real Data
Safety is the most scrutinized aspect of robotaxi operations, and for good reason. Here's what the data shows, drawn from NHTSA reports, California DMV disengagement data, and operator disclosures.
- Cumulative fully autonomous miles driven (nationwide): Over 50 million driverless miles as of late 2025
- Contact rate: Waymo's internal data, validated by third-party studies (including a 2024 study by the Insurance Institute for Highway Safety), shows Waymo vehicles have a lower police-reported crash involvement rate per million miles than human-driven benchmarks in comparable urban environments
- Serious injury crashes: Extremely rare in the Waymo fleet; the vast majority of reported incidents involve minor contact (e.g., a Waymo being rear-ended at low speed, or scraping a stationary object)
- Disengagement rate: California DMV data shows Waymo's disengagement rate (times a safety driver had to take over) has steadily improved, dropping below 1 disengagement per ~10,000+ miles in recent reporting periods — and this metric becomes largely irrelevant for fully driverless operations where there is no driver to "disengage"
Important context: Robotaxis have been involved in notable incidents. The most significant in San Francisco was the 2023 Cruise incident, in which a pedestrian — already struck by a human-driven vehicle — was run over and dragged by a Cruise robotaxi. This led to the revocation of Cruise's operating permit by the California DMV and ultimately the shutdown of the company. Waymo has had its own minor incidents, including collisions with parked cars and temporary obstructions of emergency vehicles, which drew scrutiny from San Francisco's fire department and city officials.
The broader finding, supported by NHTSA's ongoing data collection, is that Level 4 robotaxis operating in geofenced urban areas appear to be at least as safe as human drivers on a per-mile basis for the types of driving they perform — but the total miles driven are still a fraction of human driving, so the statistical confidence intervals remain wide.
The Technology Stack: How Driverless Cars See and Navigate San Francisco
San Francisco's driverless cars don't rely on a single sensor. They use a multi-layered perception system that combines several technologies:
- LiDAR: Multiple roof-mounted and bumper-mounted LiDAR units creating a real-time 3D point cloud of the environment. LiDAR works in darkness, rain, and fog — critical for SF's microclimates.
- Cameras: Dozens of high-resolution cameras providing 360° visual coverage for lane markings, traffic signals, pedestrian detection, and object classification.
- Radar: Millimeter-wave radar units for velocity detection of moving objects (other cars, cyclists) and redundancy in poor visibility.
- HD Maps: Pre-built, centimeter-accurate maps of every street in the service area, updated continuously. The vehicle knows the exact position of every curb, crosswalk, and stop sign before it arrives.
- Compute: Custom-designed AI hardware processing sensor fusion, prediction, and planning in real time.
This is a fundamentally different approach from Tesla's camera-only ("Tesla Vision") system used in consumer FSD. Waymo and Zoox rely heavily on LiDAR, which provides direct depth measurement rather than inferring it from 2D images. This sensor redundancy is a key reason regulators have approved Level 4 operations for these vehicles but not for camera-only consumer systems.
Geofencing is another critical element. Waymo vehicles only operate within a precisely defined geographic boundary. The vehicle's system has been trained and validated on every street within that boundary. If you request a ride to a destination outside the geofence, the app will not allow the booking.
Current Limitations: What Robotaxis Cannot Do
- Heavy fog and rain: While LiDAR performs better than cameras in low visibility, extreme fog (common in SF's western neighborhoods) can still degrade sensor performance. Waymo may reduce service or restrict certain routes during severe weather events.
- Construction zones: Rapidly changing road layouts, temporary barriers, and flaggers can confuse the system. Waymo may route around construction zones or, in rare cases, pull over and request remote assistance.
- No highway driving (currently): Waymo's SF operations are focused on surface streets. Highway-speed autonomous driving in mixed traffic is technically more challenging and not yet part of the San Francisco service.
- Geofence boundaries: You cannot ride a Waymo from San Francisco to Oakland, Marin, or the South Bay. The service is confined to its approved operational area.
- Complex unprompted scenarios: Unusual events — a police officer directing traffic with hand signals, a parade route, an accident scene with emergency vehicles blocking the road — can cause the vehicle to stop and wait for remote guidance or find an alternate route.
- No car seats or accessibility accommodations: Waymo vehicles do not provide child car seats, and while they are working on wheelchair-accessible vehicles, the current Jaguar I-PACE fleet requires passengers to enter a standard vehicle cabin.
- Wait times: Depending on demand and your location within the geofence, wait times can range from 2 to 15+ minutes. Outlying areas of the service zone may have longer waits.
The Regulatory Landscape: Who Approves Robotaxi Operations?
Operating a driverless taxi in San Francisco requires approval from multiple regulatory bodies, making California one of the most tightly regulated AV environments in the world:
- California DMV: Issues the Autonomous Vehicle Testing and Deployment permits. The DMV can suspend permits for safety violations, as it did with Cruise in October 2023.
- California Public Utilities Commission (CPUC): Regulates the commercial ride-hailing aspect. CPUC approval is required before a company can charge passengers for driverless rides. Waymo received its CPUC commercial deployment permit in 2023.
- NHTSA (federal): Oversees vehicle safety standards at the federal level. NHTSA has issued standing general orders requiring AV operators to report crashes involving their vehicles. The agency can investigate and order recalls.
- San Francisco city government: While the city does not directly approve robotaxi permits (that's a state function), SF officials — including the Board of Supervisors and the SF Fire Department — have exerted significant influence through public pressure, formal letters, and operational agreements with operators regarding fleet size, parking behavior, and emergency vehicle interaction.
This multi-layered oversight is unique. In states like Arizona (Phoenix), regulation has been considerably lighter, which is why Waymo launched there first. California's approach is more cautious but also generates more public data, which benefits researchers and consumers trying to assess safety.
How to Actually Use Waymo in San Francisco
If you're visiting or living in San Francisco and want to try a driverless ride, here's the process:
- Download the Waymo One app (available on iOS and Android).
- Create an account and add a payment method. There is no longer a waitlist — the service is open to the general public.
- Enter your pickup and drop-off locations. The app will confirm whether both are within the service area.
- Review the price (quoted upfront, no surge pricing) and confirm the ride.
- Meet the vehicle. The app shows the car's location in real time. When it arrives, you'll unlock it via the app by tapping a button when you're next to the car.
- Start the ride. Once you're buckled in, tap "Start Ride" on the in-seat screen. There is no driver to talk to — the vehicle departs autonomously.
- During the ride, you can watch the vehicle's perception display (showing what the sensors "see" — other cars, pedestrians, cyclists rendered in 3D). You can also contact Waymo support via the app or in-vehicle button at any time.
The experience is remarkably uneventful — which is the point. The vehicle drives conservatively, obeys speed limits, signals every lane change, and comes to complete stops at stop signs. For first-time riders, the initial minutes can feel surreal. By the end of the ride, most passengers are checking their phones.
Frequently Asked Questions
Can I hail a driverless car anywhere in San Francisco?
Within Waymo's geofence, which covers most of San Francisco proper as of 2026, yes. However, some far-outlying neighborhoods or areas near the geofence boundary may have limited availability or longer wait times. Rides to SFO airport are supported. Rides across the Bay Bridge to Oakland or south to Daly City may or may not be available depending on the current service boundary.
Are robotaxis cheaper than Uber?
Generally, Waymo pricing is comparable to UberX and Lyft Standard — within a few dollars for most trips. The key advantage is price consistency: Waymo does not apply the extreme surge multipliers that Uber and Lyft use during peak demand. For airport runs, Waymo is often slightly cheaper than a surged Uber but may be slightly more than an off-peak Lyft.
Have there been any serious robotaxi accidents in San Francisco?
The most serious incident was the October 2023 Cruise event, in which a pedestrian was run over and dragged by a Cruise vehicle after being struck by a human-driven car. This led to Cruise's permit revocation and eventual shutdown. Waymo has not had a comparable serious injury incident in San Francisco, though it has had minor collisions (typically low-speed contact with stationary objects or being rear-ended by human drivers). NHTSA and the California DMV continue to monitor all incidents.
What happens if the robotaxi encounters a situation it can't handle?
The vehicle will pull over safely and either wait for the situation to resolve (e.g., a temporary obstruction clears) or contact a remote assistance operator who can provide guidance. The remote operator does not "drive" the car remotely — they provide high-level instructions (e.g., "it's safe to proceed around the obstacle"), and the vehicle executes the maneuver autonomously.
Can I use a robotaxi if I use a wheelchair?
Waymo's current fleet (Jaguar I-PACE) is not wheelchair accessible. Waymo has announced plans to introduce wheelchair-accessible vehicles in the future, but as of early 2026, passengers must be able to enter a standard vehicle cabin independently. Zoox's purpose-built vehicle was designed with accessibility in mind, but Zoox's public service availability in SF remains limited.
Is Tesla's FSD the same as a Waymo robotaxi?
No. Tesla Full Self-Driving (Supervised) is a Level 2 driver-assist system. A human driver must be in the driver's seat, paying attention, and ready to take over at any moment. Waymo is Level 4 — fully driverless within its service area, with no human driver present. The technology, regulatory approval, and safety validation processes are entirely different.
