A navigation structure usually fails not because it is illogical but because its logic is not the reader's. This prompt reads interview language for the shape people already carry: how they group things, whether they think in tasks ("I need to change my address") or topics ("I need information about accounts"), which relationships they expect to find, and which words they use for things you have named something else. That last output, the mapping of their terms to yours, is often the one you can act on the same afternoon.
It needs transcript language to work, and enough of it that a pattern can be told apart from a personality. Say who your participants are, because mental models differ sharply with experience: a first-time visitor and someone who uses the site weekly are describing two different structures, and averaging them produces one that serves neither. Read the output for its conflicts rather than its consensus. The section where models disagree usually tells you more about the decision in front of you than the dominant model does, because it is where you actually have to choose.
You're not looking for the structure people described. You're looking for the structure their words imply they already expect.
Fill in your own details below; the prompt updates as you type. When it's ready, copy it into Claude or whatever AI tool you use.
Reads interview language for how users group concepts, name things, and expect to navigate.