How CarTold Uses AI

Where AI fits

For a model-year summary, we assemble a data package containing the relevant recall records, complaint breakdowns, investigations, manufacturer communications and fuel economy figures. The model is asked to explain that package in plain English. Its own background knowledge is not accepted as a source for a vehicle claim.

AI may also assist with drafting and editing guides, research and site text. Those pieces still require source checks and editorial review. AI assistance is not hands-on automotive experience or expert review.

What we check

The model-year pipeline checks numbers against the input package, scans for unsupported verdicts and flags similar text. A number appearing in the input is not enough: it can still be used in the wrong sentence or refer to the wrong group of vehicles. A human reviewer must check the meaning, scope and attribution.

Generated drafts are held for review. An update to a published summary stays separate from the live text until it is approved. Data tables can refresh through imports without regenerating that summary.

Claims we will not publish without evidence

• Invented facts, figures, quotations or experience with a car.
• A diagnosis or prediction about an individual vehicle based on complaints about its model year.
• A claim that fewer complaints make one year safer or more reliable.
• A free-repair promise based only on the existence of a service bulletin.
• A claim that a test drive, VIN search or report proves a car has no problems.

If we get something wrong

Send us the page and the wording you question through our corrections process. An AI-assisted sentence gets the same scrutiny as any other published claim. See our editorial policy for the standards applied to all our work.