Check the Claims in an AI Answer

Below is an answer produced by an AI assistant. I intend to rely on parts of it. Help me decide which parts to verify first. You are reviewing the answer, not writing a new one.

What the answer was for: {{purpose}}
How costly an error would be: moderate
What I already know to be true: nothing in particular

Answer to check:
{{answer}}

Do this:
1. Extract every distinct factual claim, number, name, date, quotation, citation and recommendation. Number them. Do not extract opinions or generalities.
2. For each claim, classify it as one of: Checkable in a primary source (law, documentation, official data, a paper), Checkable by running or testing something, Reasoning (follows from other claims), Opinion or judgment, or Unverifiable from what is given.
3. For each claim, rate your confidence that it is correct as High, Medium, Low or Cannot tell, and say why in a few words. Be honest that you cannot browse or confirm anything from this conversation unless I have given you sources, and that your confidence reflects pattern-matching, not proof.
4. Flag claims that look fabricated or suspiciously precise: references, URLs, quotes, statistics with no source, version numbers and legal or medical specifics.
5. Contradictions: list any claims that conflict with each other or with the facts I gave you.
6. Give me a verification order, starting with the claims where an error would cost me most given the stakes. For each, say what kind of source would settle it (not an invented URL).

Do not mark a claim as verified. You can only tell me what to verify and how.

2 required fields are empty and will stay as a {{placeholder}} in the copied text.