A safety authority shows you a recall notice as it stands today. It does not show you that a sentence in it used to read differently, and it does not keep the old one.
We read each notice when it is published, keep a fingerprint of its defect, consequence and remedy text, and read it again. So we can answer a question nobody else can: whether the wording changed after the announcement.
What it can answer
- A campaign number: the one printed on the letter. Whether that notice has been reworded since publication, when we first read it, and when we last did.
- A model: Ford Explorer, Grand Cherokee, Model 3. How many notices name it, and how many of those were later changed.
- A maker: how many notices they have filed, and which of them were reworded afterwards.
What it will not do
It is not a chatbot and there is no model behind it. It matches your words to a campaign number, a model or a maker, and answers from the record. When it matches nothing it says so, because a confident wrong answer about a safety recall is worse than no answer.
It will not tell you which part failed or what the repair involves. We are still reading the notices for that, and an answer drawn from part of the record while sounding like the whole of it is exactly the fault this site exists to document. When the reading is finished, the question becomes answerable.
Where the answer comes from
Our own readings of the authority notices, not a summary of them, and not a copy of anyone else’s database. Every campaign we hold has a page carrying the notice as filed, and any later change to it, with the dates we read each version.
If you want the underlying data rather than an answer, it is published in full, and the changes go out as a feed.