Research Methodology & Trust Center

Our safety ratings and case studies are built exclusively on verifiable public data. We do not use proprietary algorithms to "score" vehicles, nor do we accept non-public datasets from manufacturers that cannot be independently audited.

Primary Sources

• NHTSA Standing General Orders: Federal reporting requirements for ADS (Level 3-5) and ADAS (Level 2) equipped vehicles.
• State DMV Reports: Specifically California, Texas, and Arizona, where the majority of AV testing and commercial operation occurs.
• NTSB Investigations: Independent federal investigations of high-profile autonomous vehicle incidents.
• Peer-Reviewed Academic Research: For establishing baseline human driving statistics and evaluating sensor efficacy.

How We Evaluate

We separate our analyses strictly by SAE Automation Level. It is intellectually dishonest to compare Level 2 driver-assistance systems (which require constant human supervision) directly with Level 4 autonomous systems (which operate without a driver in their ODD).

When looking at crash rates, we focus on miles per incident and evaluate the damage profile (e.g., rear-end vs. front-end damage) to better understand fault, as many Level 4 vehicles are rear-ended by human drivers while operating safely.

While we strive for accuracy, our readers must understand the limitations of the current autonomous vehicle data ecosystem:

• The "Human Baseline" Problem: Comparing AV miles to human miles is difficult because human fender-benders are severely underreported to authorities, whereas AV operators are often required to report even minor contact.
• Differing ODDs (Operational Design Domains): A Level 4 robotaxi operating in complex urban environments (like San Francisco) will naturally encounter more hazards per mile than a Level 2 highway-assist system, making direct mileage comparisons flawed.
• Reporting Delays: NHTSA and DMV data can lag by months. Our pages reflect the most recently available verified data, but this may not represent real-time performance.
• Incident vs. At-Fault: A reported "incident" does not inherently mean the autonomous vehicle caused the crash. We attempt to contextualize incidents where data is available, but raw numbers do not equal liability.

Because the autonomous vehicle landscape moves rapidly, static safety claims become quickly outdated. We maintain the following update cadence:

• Safety Ratings: Reviewed and updated quarterly as new NHTSA and state DMV data is released.
• Case Studies: Published as major NTSB reports conclude or significant incidents occur. Existing case studies are updated if new evidence alters the fundamental takeaways.
• Policy & Ethics Guides: Reviewed annually to reflect the latest regulatory shifts and public sentiment.

Every page utilizing our data features a clear "Source-linked update" date at the top, ensuring you know exactly when the information was last verified against public records.