What it does
Turns an evidence pack into a scored, prioritised angle map — claims with hypotheses, evidence trails, counter-arguments and arithmetic somebody else can audit. An angle is a CLAIM. "Problem-led" is an angle; "UGC talking head" is a format; "you have been doing this wrong" is a hook. It stays on claims, because mixing the three is how a test that was supposed to compare messages ends up comparing production styles. Every angle is scored out of 100 with its arithmetic shown, so the number can be argued with rather than believed. Proven signal comes from how long competitors have run that claim; evidence strength from how many customer quotes support it and how close they sit to purchase; and anything unsupported by the landing page, blocked by compliance, or requiring proof the brand does not have is removed before scoring and recorded with its reason — because a rejection is information the hook stage needs. It has no web access by design. It cannot fetch new evidence, only process what was observed and pasted in. That restriction is the point: it makes invention impossible rather than merely discouraged.
Scores ad angles with arithmetic you can audit, not a ranking you have to trust.
Example prompts
- Here are the evidence pack and the brief — build the angle map.
- Score these angles and show me the arithmetic.
- Which of these claims are blocked by the landing page or compliance?