Are We Modeling but Not Measuring VMT for AVs? Reframing the AV–VMT Debate Through Operational Evidence

A growing body of literature argues that autonomous vehicles (AVs) will substantially increase vehicle miles traveled (VMT), contributing to congestion, induced demand, and environmental externalities. Recent work by Naz and Mattingly (2026) synthesizes 25 studies and concludes that AV deployment is associated with an average increase in VMT across both shared and non-shared AV scenarios. However, the underlying literature remains dominated by simulation-based approaches, synthetic demand modeling, and legacy travel survey data collected prior to widespread rideshare adoption and operational AV deployment. This paper critically reassesses the methodological foundations of the AV–VMT literature through a structured inventory and classification of the studies included in Naz and Mattingly’s meta-analysis. The analysis identifies recurring endogenous mechanisms embedded within the literature, including reductions in generalized travel cost, travel burden, and value of travel time (VOTT), which structurally privilege AV adoption and systematically inflate projected VMT outcomes. At the same time, few studies incorporate observed AV rider behavior, empirical occupancy rates, operational pricing dynamics, or adaptive fleet learning. Drawing on emerging evidence from operational AV systems, including Waymo fleet data, shows that he share of empty VMT declined from ~63% to ~44% since 2023 for real world deployments while passenger serving VMT increased from 37% to 56%. This underscores an argument that AVs should be understood not as static hypothetical vehicles, but as adaptive operational mobility systems embedded within platform, pricing, and fleet management constraints. The paper concludes by proposing a minimum operational data standard for future AV–VMT research and argues that the field must transition from hypothetical scenario modeling toward empirically grounded operational evaluation.