# Mileage by listed condition

This additional analysis was calculated September 8 from the original September 7 captures. It includes all 1,458 sampled VDPs and adds no websites. Condition follows explicit visible labels on 1,286 pages, vehicle breadcrumbs on 21 and document titles on 151. Model year alone does not determine new or used.

| Listed condition | Sampled pages | Present | Missing | Unknown | Evaluable |
|---|---:|---:|---:|---:|---:|
| New | 956 | 509 | 445 | 2 | 954 |
| Used, including CPO | 502 | 336 | 165 | 1 | 501 |
| Total | 1,458 | 845 | 610 | 3 | 1,455 |

Missing mileage markup occurs on 445/954 new pages (46.6%) and 165/501 used pages (32.9%). New and used pages come from 476 and 481 websites respectively; those groups overlap and must not be added.

The split reconciles to the original 845/1,455 mileage-presence result. A separate raw-property census agrees on all 1,458 page-level presence flags. Thirteen supplementary direct AI-assisted source reviews comprised three distinct websites in each new/used by present/missing group, plus one used-page example. The 12 category examples used fixed hash selection.

The [mileage summary](/insights/dealership-structured-data-study-2026/assets/mileage-by-condition.csv) reports the new/used totals. Row-level records, raw mileage values, VINs and URLs remain private. [Reviewed examples](/insights/dealership-structured-data-study-2026/assets/verification.md#mileage) explain why missing markup and missing visible information are different.

“New” is a condition label, not an odometer reading. Actual known mileage and units help distinguish differently used vehicles. Unknown mileage must not be filled with zero. This is an information-quality recommendation, not a universal Schema.org requirement.