Method
Check the draft against the outline, question by question
Put the outline and the draft side by side and fill in one row per question: question, the point in the draft that answers it, and the result. If you can't name a specific point, the result is absent, not almost.
Put the outline and the draft side by side and fill in one row per question: question, the point in the draft that answers it, and the result. If you can't point to a specific passage for a question, the result is absent, not almost. If you were looking for how AI visibility is measured, that is a different step; here you check the text before you publish. This check covers only what is written in the draft, not how an engine will return it.
The method, as a working proposal from RankAI
The grid we propose is not a universal standard: it is a way of working that makes visible, row by row, where the draft answers and where it does not. Three results are enough: present when a point answers the question directly; partial when the point touches the question but leaves part of it uncovered; absent when there is no point to name. The difference between partial and absent is the most useful one, because partial can be fixed, while absent means you have to write.
Two filled-in rows as an example
Question: "How do you check that the draft answers the questions in the outline?" — Answer point: the paragraph describing the question/point/result grid — Result: present.
Question: "What happens if a visibility figure is unavailable?" — Answer point: the sentence saying the product shows N/A — Result: partial, because the point names the case but does not connect it to the draft check.
Why the result is not the same as measurement
The grid says whether the text answers, not how visible it will be. Anyone looking for how AI visibility is measured should know that is a different question: we do not cover it here. One useful point to keep in mind is that what comes back from an engine through an API is the API response, not the screen a person sees. That is why checking the draft stays a check on the text, not a preview of what someone will see. When a visibility figure is unavailable, the product shows it as N/A, never as zero, and that is a measurement rule, not a draft rule.
The contribution of this piece is the three-result grid applied to the draft, with one rule for filling it in: no row can stay empty and no result can be "to be seen." If the answer point cannot be found, write absent and move to the next row; the fix comes later, not while you fill it in. This avoids two common mistakes: marking a question present just because the topic is mentioned, and marking a question absent because the answer is spread across several points. In the second case you list all the points and the result stays partial until one is added that closes it. The grid is filled in once, before publishing, and stays as a record of what was checked.
The limit to keep in mind
The grid checks the draft, not an engine's response: what comes back through an API is the API response, not the screen a person sees, and this check does not close that gap. For the boundary between what we see and what a user sees, the reference is the explainer page, where the limit is described starting from the difference between the API response and the screen.