Where the evidence comes from

• We read the terms first

  Before we read a site for the first time, somebody reads its robots file and its terms, and records what they say and on what date. That record is stored against the source, not in someone's memory.

• We use the front door wherever there is one

  Where a source publishes an interface for the purpose, we use it rather than scraping around it.

• We are slow on purpose

  Rate limiting, exponential backoff, and a breaker that stops entirely after repeated refusals. If a source starts refusing us, the correct response is to stop asking, and that is what the code does.

• We store references and structure, not copies

  We keep what we derived and a link back to the original. We are not a mirror of anyone's content.

• We attribute on every surface

  Every review shown on an evidence page carries its source.

• Authors are pseudonymous

  We store a stable opaque key so we can spot one account posting in bursts. We do not store names, emails or profile links.

• If terms forbid it, we drop the source

  And write down why. The product is built to survive that, which is what makes the rule credible rather than decorative.

We are not fetching anything while you use the site. Every page you see was computed overnight. We are not reselling anyone’s reviews, we do not republish them in full, and we do not accept payment from any manufacturer or dealer. That last one is precisely why we can afford to say that a car’s service experience is poor.