At 7:42 on a Tuesday morning, a Waymo Jaguar pulled away from a Phoenix curb, stopped for a delivery van, and waited through a full pedestrian crossing cycle. That is the useful part of researching waymo locations: not the map graphic, but the operating boundary where the system has actually been tested. The short version is that Waymo has operated commercial ride-hailing service in parts of Phoenix, San Francisco, Los Angeles, and Austin, with additional expansion and partnership markets changing over time. A city name is not a guarantee that your pickup address is eligible.
The current operating pattern
Waymo does not treat an entire metropolitan area as one giant geofence. It launches in defined service areas, usually with restrictions tied to roads, neighborhoods, operating hours, airport access, weather, and rider eligibility. The Waymo One app is the source of truth for a specific trip. A neighborhood that appears inside a broad public map can still return “service unavailable” when the pickup pin lands outside the active polygon.
Phoenix was the early large-scale reference market. Waymo built extensive experience there across the East Valley and central Phoenix, including areas around Chandler, Tempe, Mesa, and Scottsdale. Availability has never meant every street in those cities. Construction, private roads, unusual access controls, and temporary road closures can all remove a trip from the bookable area.
San Francisco is a harder benchmark. Hills, dense curb activity, cyclists, double-parked vehicles, transit corridors, and frequent construction create a more demanding planning environment than a wide suburban arterial. When comparing waymo locations, I would separate “commercially available” from “equally difficult.” A service can be available in both Phoenix and San Francisco while facing very different perception and prediction problems.
What the major markets test
San Francisco is valuable because the vehicle must negotiate a high-entropy urban scene. A delivery worker stepping around a truck, a bus stopping outside its usual position, or a cyclist changing lanes can create more interaction branches than a quiet residential street. The correct question is not whether the vehicle completed a ride. It is whether it selected a reasonable trajectory without excessive hesitation, unnecessary yielding, or a remote-assistance intervention.
Los Angeles adds scale and road complexity. The service area has included parts of the city and nearby communities, but boundaries have expanded in stages rather than appearing all at once. Freeways, multilane arterials, curbside pickup pressure, and event traffic make Los Angeles a useful place to observe how the stack handles incomplete information. A smooth trip at noon does not prove equivalent performance during evening congestion.
Austin is another important market because it combines a growing urban core with suburban roads, wide intersections, construction zones, and different curb patterns from California. Public availability has changed as Waymo expanded its presence and partnerships. Before planning around a ride, open the app, enter the exact pickup and destination, and confirm that the vehicle type and trip are offered. Press coverage is a poor substitute for that check.
New markets and the partnership problem
The list of waymo locations also depends on what “Waymo service” means. A market can have Waymo-branded rides booked through Waymo One, rides requested through a partner platform, testing vehicles without public rides, or announced future service that is not yet generally available. Those categories are not interchangeable.
Waymo has announced or pursued expansion beyond its established markets, including Atlanta through a partnership with Uber. That does not mean every Atlanta resident can request a car at any time. Launch schedules, service polygons, fleet availability, and booking-platform rules determine practical access. Miami has also been discussed in expansion plans, but an announced market should be treated as planned until riders can successfully complete a trip in the official application.
This distinction matters for testing. If I record a result from a public ride, I need to record the city, date, weather, route type, time of day, and whether the booking came through Waymo One or another platform. Otherwise, “Waymo drove here” is not a reproducible claim. It is a map pin with missing metadata.
How to verify a ride before you travel
Use this five-step check before relying on any of the waymo locations for an airport run, appointment, or scheduled pickup.
- Update the Waymo app. Old app versions can display stale availability or lack the latest booking flow.
- Set the pickup pin on the actual curb, not at the center of a neighborhood or business complex. Large campuses often have restricted access points.
- Enter the destination and inspect the estimated pickup time. A valid destination with no car estimate can indicate limited fleet availability or a temporary service condition.
- Check the ride instructions. The app may direct you to a designated pickup zone rather than the point closest to your GPS location.
- Have a backup plan. A human-driven rideshare, taxi, or transit connection is sensible when the trip is time-critical.
The app may also limit service during severe weather, active emergencies, major road closures, or unusual event traffic. That is not necessarily a model failure. It can be a deliberate operational constraint. The engineering question is whether the constraint is visible to the rider and whether the system fails safely when the route changes after booking.
What I would measure in each city
I would not rank waymo locations by marketing footprint. I would build the same test matrix in each market: unprotected left turns, curbside loading zones, lane changes near cyclists, four-way stops, construction detours, school-zone behavior, and pickup precision. Each scenario gets a severity score from 1 to 5. A score of 1 is an awkward but harmless pause. A score of 5 is a safety-critical decision requiring immediate human intervention.
For every run, I would save the route, timestamp, weather, map context, intervention notes, and video where permitted. Then I would compare not just completion rate, but delay, unnecessary yielding, trajectory smoothness, and repeatability. I ran the same kind of route three times in simulation once and got three different answers after changing only the initial traffic state. That is why one impressive ride tells me less than a controlled set of ordinary rides.
The practical answer
The most accurate list of waymo locations is always two lists: markets where Waymo has publicly operated, and streets where the app will accept your specific trip today. Phoenix, San Francisco, Los Angeles, and Austin are useful reference points, while newer partnership markets require closer verification. Treat announcements as leads, not access confirmations.
Your car talks. I check his homework. If you want to test the system rather than admire the branding, capture the route details, document the edge case, and report what the vehicle should have done versus what it actually did. That gap is where the useful information lives.