Start with a real reader question, build a traceable source pack and define what your article adds. Search volume alone does not establish usefulness.
A list of generated topic ideas is a starting point, not evidence that readers need those articles. Choose a question you can answer more concretely than a generic overview.
Collect real questions
Look at support requests, product usage questions and available search-query reports. Separate an observed question from an inferred audience need. Low measured search volume does not prove a topic has no value; a narrow support problem may still deserve documentation.
Define a testable contribution
Write one sentence describing what the page will add. For example: “Show whether the gap is a blank line in the text or paragraph spacing in the destination editor.” This is more precise than “everything about AI formatting.”
Try the task yourself
Use a non-sensitive sample, record the input and inspect the result. If you cannot reproduce the behavior, label the uncertainty instead of building the article around an assumption.
Build a source pack
Prefer original documentation or research for technical and statistical claims. For each source, keep the URL, date checked, supported claim and limitation. Community discussions can reveal questions, but an anecdote does not establish a general failure rate.
Open every citation suggested by an AI tool. Check whether the quoted material exists and says what the draft claims.
Create the brief
Include the reader, question, opening answer, worked example, relevant sources and claims to avoid. Ask the draft to preserve uncertainty rather than fill gaps with plausible numbers.
Before publication, use Google's people-first review questions as a second perspective, not a guarantee of traffic.
For the example above, compare the spacing guide with the spacing cleaner. Link to a tool only when it helps complete the reader's task.