Frequently Asked Questions
Quick answers to common questions about this project and the data behind it. For a more detailed technical discussion, see our Methodology page.
Two public sources: crash reports Waymo files with the National Highway Traffic Safety Administration (NHTSA) under its Standing General Order, and Waymo's Safety Impact Data Hub.
The federal reports are the backbone: every crash on this site starts there. Waymo's quarterly data release then adds exact dates, street locations and crash-type classifications, plus the miles Waymo has driven. Because Waymo's release lags the federal data by a few months, the newest crashes appear with month-level dates and approximate locations until the next release fills them in. Details are on the methodology page.
Whatever Waymo is required to report. Since June 16, 2025, that means any crash involving a death, a hospital-treated injury, an airbag deployment, a pedestrian or cyclist, or a towed vehicle; plus any other crash where damage is expected to exceed $1,000 or where the Waymo hit another vehicle or object. Before that date, any crash with any property damage counted — including very light bumps.
So the dataset is full of minor incidents: most involve no injury at all. That's a strength (it's far more complete than police data for human drivers) and a reason to read raw counts carefully.
Yes. On June 16, 2025, NHTSA's Third Amended Standing General Order took effect. It stopped requiring reports for many minor crashes (roughly: the self-driving car was hit and damage stayed under $1,000), extended the deadline for the most serious reports from one day to five, and stopped requiring every company involved in the same crash to file.
The effect is visible in the data: Waymo's rate of all reported crashes per mile fell sharply across its cities at the same time, while its rate of injury crashes — reportable under both rules — barely changed. Comparisons across June 2025 are not like-for-like.
The data runs from -- through --, but most crashes are from 2024 onward, as Waymo grew from two cities to fifteen.
The site updates itself every Monday. Federal reports currently run through --, but the latest months are still filling in — companies have until the 15th of the following month to report most crashes, and NHTSA publishes with a further lag — so we consider the record complete through --. Waymo's detailed data runs through --.
Headline numbers cover driverless operation only. Crashes during supervised testing with a safety driver (-- on record) can be shown in the Explore map.
Most are minor. Of -- reported driverless crashes, -- involved a moderate, serious or fatal injury to someone, and -- involved a death. Among crashes Waymo's detailed data covers, --% involved any injury.
According to the reports Waymo filed and to police accounts, in each fatal crash summarized below the Waymo was stopped or slowing and was hit as a result of another road user's actions; no Waymo has been found at fault:
- San Francisco, January 2025: an empty Waymo stopped in traffic was hit when a speeding driver crashed into a line of cars. One person and a dog were killed; the speeding driver was charged.
- Tempe, Ariz., September 2025: a motorcycle hit an empty Waymo from behind as it slowed to turn; the rider was then run over by a passing car whose driver fled and later faced a felony charge.
- Dallas, August 2026: an SUV struck a pedestrian, who was thrown into an empty Waymo that had slowed to about 5 mph; police had not publicly assigned fault in the latest reports.
On the map, crashes are colored by severity. Click a marker to read the narrative Waymo filed.
In --% of crashes the Waymo was stopped or parked, and in --% it was going 5 mph or slower. The most common crash is a rear-end collision (--% of classified crashes); in --% of those, the Waymo was the car hit from behind. Side-impact (--%) and backing collisions (--%) come next.
Not necessarily. Federal reports don't assign fault, and neither does this site. A Waymo stopped in the middle of an intersection, or braking unexpectedly, can contribute to a crash it doesn't physically cause. The facts here describe what happened, not who was responsible.
Hard to say yet. Waymo's rate of all reported crashes per mile fell by almost half in mid-2025 — but that coincides exactly with the reporting-rule change and appears across its cities at once, so much of it reflects paperwork rather than driving. Its rate of injury crashes, which is unaffected by the change, has stayed roughly flat at well under one per million miles. See the charts on the home page.
Waymo publishes comparisons with human drivers on the same kinds of roads in the same cities. Over -- driverless miles through --, it reports -- fewer serious-injury-or-worse crashes and -- fewer injury crashes. Caveats:
- It's Waymo's own analysis, though the method has been peer-reviewed (Kusano et al., 2025).
- Human benchmarks can't perfectly match where, when and how Waymo drives.
- The most severe outcomes rest on few crashes, so their ranges are wide — the home page shows the uncertainty for each.
An independent study by the Insurance Institute for Highway Safety (July 2026), covering about 50 million miles from 2021–2024, found 68% fewer police-reportable crashes and 81% fewer injury crashes per mile than human drivers — with a much smaller gap in San Francisco (35% fewer) and slightly more crashes in Austin, where its sample was small.
As of October 2026, Waymo carries public riders in -- metro areas: Phoenix (since 2020), the San Francisco Bay Area (2023), Los Angeles (2024), Austin and Atlanta (2025), and — all in 2026 — Miami, Dallas, Houston, San Antonio, Orlando, Nashville, Denver, San Diego, Tampa and Las Vegas. Some of the newest are open to a limited group of riders at first.
It is testing without anyone in the driver's seat in Detroit, and with safety drivers in Washington, D.C. and Philadelphia, where local rules don't yet allow empty cars. It has announced plans for several more U.S. cities, including New York, Chicago and Seattle, and abroad.
Driverless crashes on record by metro area:
The San Francisco Bay Area includes Waymo's Peninsula and San Jose service area. Raw counts mostly reflect how much Waymo drives in each place — see the per-mile rates below.
Through --, -- driverless miles in all. Waymo publishes city-level figures for its longest-running markets:
Differences between cities likely reflect traffic density and how much of each city's driving happened under the older, broader reporting rules — compare with care. Waymo hasn't yet published mileage for its newer cities.
Crashes peak in the afternoon and evening and on Fridays and Saturdays; --% happen during rush hours (7–10 a.m. and 5–8 p.m.). That mostly reflects when roads are busiest and when Waymo's cars are out carrying riders. Waymo doesn't publish miles by hour, so the timing alone can't tell you whether some hours are riskier.
Not all of them, and each marker says how precise it is. Federal reports redact addresses; the street descriptions come from Waymo's quarterly release. Of the driverless crashes, -- are placed at the intersection or address, -- on the right street near it, -- at the center of their ZIP code, and -- — mostly the newest crashes — near the center of their city. Markers upgrade automatically when Waymo's next release covers them.
Freshness: Waymo's hub is updated quarterly and lags a few months, while this site reads the federal record weekly, so recent crashes appear here first. For the periods both cover, the two agree almost one-for-one — a useful check on both. NHTSA's data also contains -- Waymo crashes from supervised testing, which Waymo's hub leaves out by design and which we keep out of the headline numbers too.
Yes. Crash counts alone don't capture that kind of scrutiny, so here is the record as of October 2026. A "recall" for a self-driving fleet usually means a software update sent to every car, not a trip to the shop.
- School buses (2025–26). After NHTSA opened an investigation (PE25013, October 2025) into Waymos passing stopped school buses, Waymo recalled about 3,000 vehicles in December 2025. Austin's school district reported further violations after the fix, and the National Transportation Safety Board opened its own investigation in January 2026.
- Child struck near a school (January 2026). NHTSA (PE26001) and the NTSB investigated after a Waymo hit a child who ran into the street from behind an SUV near a Santa Monica elementary school. The car braked from 17 mph to under 6 mph; the child's injuries were minor.
- Floodwater (April–May 2026). After an empty Waymo drove into a flooded San Antonio road and was swept into a creek, Waymo recalled about 3,800 vehicles. San Antonio service was paused from April until mid-September 2026.
- Freeway work zones (June 2026). Waymo recalled 3,871 vehicles after its cars entered closed freeway construction zones 13 times in Phoenix and the Bay Area; no one was hurt. Freeway service was restricted while a fix was developed.
- Earlier. Three smaller software recalls in 2024–2025, including one after a Waymo hit a utility pole.
Yes. Download every crash as a CSV file (rebuilt weekly; columns are explained in the data dictionary). The original sources are public too: NHTSA's Standing General Order data and the Waymo Safety Impact Data Hub. The code that builds this site is open source.
Anders Eidesvik and Kate Li, as a data journalism project begun in Stanford's COMM 277T: Building News Apps. See the About page. Corrections are listed on the methodology page.