Which sources does SpecAxle use?
SpecAxle uses original equipment manufacturer material whenever it is available. Examples include owner manuals, towing guides, technical data books, service documents, wheel-fitment bulletins, and official product pages.
SpecAxle uses the NHTSA Vehicle Product Information Catalog, or vPIC, to build the make, model, model-year, trim, and engine identity skeleton. A vPIC identity does not become a towing, payload, or GVWR value unless an allowed source supports that value.
How does the verification workflow work?
- Register the source. SpecAxle records the title, publisher, URL, document year, and file checksum.
- Match the vehicle. The source must identify the relevant year, make, model, trim, engine, drivetrain, and required equipment.
- Extract candidate rows. Deterministic table extraction is preferred for structured documents.
- Verify the evidence. Each candidate must retain its unit, conditions, page reference, and exact supporting text when text extraction is used.
- Run quality gates. Range checks, configuration checks, and cross-source comparisons reject unsafe or conflicting rows.
- Publish the record. The value and its provenance enter the database together.
- Record the date. The visible “Last verified” date changes only after the source record is checked.
Can AI extract a candidate specification?
SpecAxle may use a language model to extract a candidate from difficult document text. The model receives the source text and a known vehicle list, uses temperature zero, and must return an exact quote plus page number for every value.
SpecAxle automatically rejects a candidate when its quote does not appear in the source text. AI-extracted candidates pass the same validation and conflict checks as deterministic table rows. An AI model cannot create, estimate, or complete a missing value.
What happens when two sources disagree?
SpecAxle flags conflicting values for human review and does not choose a winner automatically. The reviewer checks model year, market, configuration, units, package requirements, document revision, and source authority before publication.
How does SpecAxle correct an error?
Every specification page includes a “Report an error” link. A correction report enters the same source-verification workflow as a new record. SpecAxle changes the published value only when evidence supports the correction, then updates the verified date and purges affected cached pages.
Send correction evidence through the SpecAxle contact page. Include the page URL, exact configuration, disputed value, and primary-source document.
How often does SpecAxle update data?
SpecAxle reviews source availability monthly and processes new model-year documents when manufacturers publish them. Corrections and revised OEM documents receive priority. The specification sitemap is rebuilt nightly, but a record’s verified date changes only after a real source review.
How should SpecAxle data be cited?
Use the clean canonical page URL and the visible verified date. The preferred format is: SpecAxle, {canonical URL}, verified {YYYY-MM-DD}. Retain the named OEM or government source when space permits.