Separate the model, the application, and the evidence behind an answer.
3 min read
The short version
When you ask an AI a question, two things are at work: the app you are typing into, and the model behind it. The app decides what the model gets to see. The model guesses a good answer from whatever it was handed. To check an answer, ask which of the two produced each part of it.
An AI answer comes from a model working inside an app. The model produces text from the input it gets. The app may give it files, search results, or tools. These are different parts of the system.
What to check
A model can write from patterns learned in training. It can also use new facts supplied by the app. You need to know which sources support this answer. A strong model can still answer from the wrong file.
Suppose an email says, ‘I sent the notice yesterday.’ That tells you what the sender claimed. It does not prove when the notice arrived. Check the email's date, the exact words, and any record of receipt.
Try it
Which answer is easier to check: a date alone, or a date with a quote and a source link?
Check your understanding +
The linked answer is easier to review. You must still open the source. Check whether it proves the claim, and whether the answer left out a limit. A source link is a way to check, not proof of truth.
Examples are teaching material. For legal rules, read the current primary text linked below. A revision date does not certify every claim.
Sources and further reading
Linked sources support the technical background identified below. The legal workflows, prompts, and fictional examples are original editorial exercises; they are not claims of endorsement by the linked authors.
