By the end of 2025, around 2,500 robotaxis were actively operating in the United States, with further expansion well underway. This study, from Houston self-driving car accident lawyer Attorney Brian White, examines where autonomous vehicles face the most demanding driving conditions, factoring in federally reported automated vehicle crashes.
It also compares driving conditions across U.S. robotaxi markets, examining fatal traffic crash rates, pedestrian and cyclist risks, and intersection safety.
U.S. Cities with the Toughest Driving Conditions for Robotaxis
As robotaxi services continue to expand across a growing number of American cities, they face an increasing range of distinct, often challenging road environments that feature varying levels of traffic danger, pedestrian activity, congestion, and risks around intersections.
Houston offers a timely test case, with Waymo offering a fully autonomous service to the city from August 20, 2026, and Zoox announcing plans to begin Houston testing on September 1.
Yet how challenging are Houston’s roads likely to be for robotaxis?
Our new analysis of federal traffic and automated-driving data compares 15 currently active robotaxi markets. It also examines where the highest proportion of reported automated-driving incidents occur, closely considers pedestrian and cyclist incidents, and looks at the regulatory frameworks that govern robotaxi operations.
The results (more details below) put Houston in third place (of 15 cities) for 2024 fatal traffic crashes per 100,000 residents, behind Phoenix and Dallas.
(Note: this ranking measures the existing road environment, not robotaxi performance. A city’s overall crash rate can’t establish whether autonomous vehicles are safer or more dangerous on its roads, and the appearance of an automated vehicle in a federal incident report doesn’t establish its cause.)
The analysis instead asks: What kind of road environments are robotaxis negotiating, and what do residents need to know as driverless services increase?
For a legal consultation call (713) 500-5000
Which Robotaxi Markets Face the Greatest Road Challenges?
The 15 robotaxi markets we examined feature a range of traffic-fatality rates.
Phoenix records the highest rate (16.06 fatal crashes per 100,000 residents), followed by Dallas (15.47). Houston ranks third (13.87), ahead of Nashville (13.35) and Tampa (13.32). Here’s the full list of 15 cities.
The national 2024 average rate was 10.68 fatal crashes per 100,000 residents, putting Houston some way above the national benchmark. Additionally, Houston’s intersection-related fatal-crash rate (3.27 per 100,000 residents) was higher than the national average (2.86). It also recorded a pedestrian death rate of 4.86 per 100,000, and a pedalcyclist death rate of 0.38.
Although these figures provide useful context for the environment robotaxis enter, they shouldn’t be interpreted as any indication of a readiness rating. The overall ranking measures all traffic (as opposed to just autonomous vehicles), and while it adjusts for resident population, it doesn’t factor in miles driven, trips, fleet size, or fault.
One of the reasons Houston represents a useful example is that the city’s robotaxi availability is evolving from limited to broadly available. Waymo made its Houston service available to everyone on August 20; for Zoox, its September Houston expansion initially involved testing and mapping as opposed to an unrestricted public robotaxi service.
Houston’s congestion rate also increased from 38.8% in 2024 to 40.8% in 2025; that said, congestion levels are not part of the fatal-crash ranking.
The recorded figures raise practical questions for all cities expanding their autonomous vehicle services. They include: are pickup and drop-off locations easy to predict? Can emergency crews quickly contact or move a stopped vehicle? How clear are operating boundaries and access rules? And is there any incident and mileage data made available so residents can assess and compare robotaxi performance?
The purpose of those questions isn’t to rank one city safer or more prepared than another; it’s to identify the information and infrastructure that will be increasingly important as robotaxi services continue to grow.
Where Reported Automated-Driving Crashes Occur Most
When our analysis moves beyond all-traffic crashes and is limited to federally reported automated-driving-system incidents, the state rankings change significantly.
Between August 2025 and July 2026, Arizona recorded the highest population-adjusted ADS incident rate of all examined states (2.138 per 100,000 residents). California followed (1.733), with Nevada (1.036) and Texas (0.829) close behind.
The national average selected-report rate was 0.369 per 100,000 residents.
These numbers shouldn’t be read as a clear ranking of the most dangerous states for autonomous vehicles. Robotaxi deployment is varied, and reporting needs often determine which incidents become part of the federal dataset. A state with more reported incidents could feature more qualifying automated vehicles on its roads.
The very specific circumstances of each reported incident also provide important context. In 51.7% of qualifying incidents, the vehicle was stopped or parked before the crash. Also, 34.9% involved a vehicle driving straight, while 6.6% involved a vehicle turning.
The assessment of injuries was also limited in most reports. 89% recorded no injuries, with 9.4% featuring injuries or fatalities, and 1.6% of unknown severity.
Those figures clarify why an incident count can be misleading. Involvement in a reported incident doesn’t confirm fault, while the circumstances that surround the event vary and often determine key details.
An independent mileage analysis offers another perspective. According to the Insurance Institute for Highway Safety, Waymo’s driverless vehicles recorded 68% fewer police-reportable crashes and 81% fewer injury crashes per mile than equivalent human-driver benchmarks between 2021 and 2024.
It should be emphasized: those two different perspectives answer different questions and can’t be combined. The federal state comparison uses residents; the IIHS comparison uses miles driven.
The dataset also features details regarding higher levels of automation, including SAE Levels 3–5 and testing activity, as opposed to Level 2 driver-assistance systems that require human supervision.
Ultimately, population-adjusted incident rates show where reported activity occurs, but can’t determine vehicle safety or operational readiness.
Click to contact us today
Where Federal Reports Involve Pedestrians
It’s also difficult to draw conclusions about pedestrian involvement, with limited pedestrian-related ADS reports available.
In fact, Arizona and California are the only states with qualifying pedestrian-category records in the analysis, which means the data can’t support a meaningful 10-state ranking.
Of the two state records available, Arizona reports 0.0131 pedestrian-category ADS incidents per 100,000 residents; California reports 0.0051.
Individual reports underscore why the term ‘pedestrian incident’ can be problematic. For example, a November 2025 Scottsdale report and a June 2026 report from San Francisco involved passengers getting out of moving vehicles: this is not a conventional type of pedestrian collision.
In Santa Monica in January 2026, a pedestrian appeared from behind an SUV and suffered a minor injury that didn’t need a hospital trip.
Realistically, pedestrian reports represent a much wider set of interactions than a vehicle striking someone crossing the road. Key incidental factors such as pickup locations, passenger exits, parked vehicles, and low visibility can all influence whether or not an autonomous vehicle hits a pedestrian.
National pedestrian data may provide useful context. 2024 Governors Highway Safety Association data suggests that 62.2% of pedestrian fatalities occurred at locations that lacked a sidewalk, while 76.5% of fatalities featuring known lighting conditions occurred after dark.
In 2024, Houston’s pedestrian death rate was 4.86 per 100,000 residents, compared with a national average of 2.08.
These national and city-level figures are not robotaxi fatalities: they merely show the street conditions autonomous vehicles need to navigate.
For robotaxi operators, the issue extends beyond avoiding collisions. Other issues include whether or not such vehicles can identify pedestrians as they emerge from behind a parked car, how far away pickup locations are from high-conflict areas, and if a particular street provides pedestrians with predictable, protected crossing points.
Federal ADS records can’t answer all those questions, partly due to the small number of qualifying pedestrian reports. However, they can illustrate why pedestrian interaction needs broader consideration than a simple collision count.
Complete a Case Evaluation Form Now
What Happens When Robotaxis Interact With Cyclists?
ADS records relating to robotaxis colliding with cyclists are also limited, but those that are available highlight several different types of interaction.
Four states feature qualifying cyclist incidents, with California recording the highest rate (0.0178 per 100,000 residents), followed by Texas (0.0095), Georgia (0.0088), and Florida (0.0043).
Individual reports establish the importance of passenger behavior and secondary collisions.
For example, in San Francisco (March 2026), a passenger opened a rear door directly into a cyclist’s path, causing a minor injury. And in Miami Beach (July 2026), a separate vehicle struck a cyclist and knocked them into a stationary automated vehicle, leading to serious injury.
In neither case does the report establish a defect or fault with the autonomous-driving system. Instead, both offer good examples of the type of interaction that can happen around a robotaxi.
The national cyclist picture adds an additional key dimension. The U.S. pedalcyclist death rate rose from 0.324 per 100,000 residents (2024) to 0.336 (2025), even though the overall number of traffic deaths fell.
In 2024, 81% of pedalcyclist deaths occurred in urban areas, with 30% occurring at intersections. The pedalcyclist injury rate also rose from 14.84 per 100,000 (2023) to 15.55 (2024).
There aren’t enough available federal ADS cyclist records to establish reliable risk estimates. But the combination of passenger-exit incidents, secondary collisions, and a rise in cyclist deaths in the U.S. underscores the importance of being able to predict where vehicles are in relation to cyclists.
For cities, that can mean understanding where robotaxis are likely to stop, how passengers are instructed to exit, and how vehicles behave around bike lanes and other tight road spaces.
The data also features a classification caveat: the cyclist category includes a pedestrian carrying a bicycle and a pedicab (the NHTSA definition has included motorized bicycles since 2022). So it doesn’t perfectly isolate conventional bike riders.
What Texas Can Learn From California’s Robotaxis
As robotaxi services increase, regulatory issues become increasingly important. But what information and safeguards can residents, emergency responders, and regulators use?
Texas and California follow different approaches when it comes to autonomous-vehicle oversight.
In Texas, mandatory authorization for qualifying commercial driverless operations became enforceable on May 28, 2026. Level 4 and Level 5 operators must have authorization and an emergency-interaction plan filed with the Texas Department of Public Safety, while TxDMV can restrict, suspend, or revoke authorization.
California’s Notice of Autonomous Vehicle Noncompliance process started on July 1, 2026. This is a formal regulatory enforcement mechanism as opposed to a conventional traffic ticket issued to an autonomous vehicle.
California also divides its authority between two different agencies. The Department of Motor Vehicles handles autonomous-vehicle tests and deployment permissions, while the California Public Utilities Commission is responsible for passenger service. The state’s four relevant programs distinguish between drivered and driverless pilots and deployments, with pilots not allowing fares and deployments allowing paid passenger service.
This contrast doesn’t mean one state’s framework gets better safety results. Instead, it illustrates how different regulatory systems can address the same emerging technology.
Houston’s ADS incident rate in the study is 0.501 per 100,000 residents. The Houston records referenced are classed as property damage with no injuries reported. The figure in question shouldn’t be interpreted as a citywide robotaxi safety measure.
One reason transparency is so important is that incident datasets can miss a significant share of events depending on reporting rules. A 2021-2024 IIHS analysis found that only about 22% of autonomous road crashes were police-reportable.
For Texas, potential improvements could include publishing usable mileage and incident data, distinguishing testing from real-world deployment, recording blocked-road or stopped-vehicle incidents, and making sure first responders can reliably contact autonomous vehicles.
Although these measures would not clarify whether robotaxis are safe or unsafe, they would give residents and regulators a better chance of assessing how the technology performs as deployment expands.
What This Means for Houston
Houston now provides a useful illustrative example regarding what happens when robotaxis enter a large, complex urban road network.
Its 2024 fatal-crash rate (13.87 per 100,000 residents) puts it third among the 15 examined robotaxi markets. Its intersection-related fatal-crash rate is also above the national benchmark, and its pedestrian death rate is more than double the national figure.
Yet Houston’s robotaxi story can’t be inferred from those traffic statistics since they describe the city and its existing roads, not how autonomous vehicles perform.
Federal ADS data provides a different picture, shaped by deployment levels, reporting requirements, and individual incident circumstances. The small number of pedestrian and cyclist records means those categories are unsuitable for broad safety rankings; nonetheless, the available cases show how passenger exits, parked vehicles, cyclists, and secondary collisions can create complex interactions.
Overall, that leaves transparency as the common factor.
As autonomous vehicles become a feature in a growing number of cities, residents need to be able to know the difference between general road danger, reported autonomous-vehicle incidents, and measured vehicle performance.
Ultimately, Houston’s challenging road environment offers an opportunity to ask more questions, not a basis for declaring autonomous vehicles as either safe or unsafe.
A useful test will be what happens as robotaxis log more real-world miles, and whether operators and regulators provide enough usable data to allow the public to evaluate the results.
Robotaxi Performance Demands Better Assessment
Clearly, robotaxi services are expanding in U.S. cities that feature different road-safety environments, with Houston currently ranking third among 15 current markets for fatal crashes per 100,000 residents. Separate federal ADS data shows that reported incidents are disproportionately featured in a small number of states, while pedestrian and cyclist records remain relatively rare.
Overall, robotaxis are entering a growing number of challenging road environments, but citywide crash rates and reported ADS incidents don’t by themselves indicate the safety or danger levels of autonomous vehicles. Comparable mileage data, transparent incident reporting, and clear oversight are needed to assess their real-world performance.
Self-driving cars are no longer an outlandish or speculative concept: they’re already on Houston roads. And in complex ways that differ from a typical car crash, accidents involving autonomous or semi-autonomous vehicles can raise difficult questions about fault. Our self-driving car accident lawyer in Houston can guide you through the legal process.
Attorney Brian White has over 20 years of experience helping injury victims. Contact us to learn about your rights and what your legal options for recovery might be. We could help recover your medical bills, the cost of property damage, lost wages, and more.
Call or Text (713) 500-5000 or Complete a Case Evaluation Form
