What the question set actually does
It serves two purposes at once: it tells you what to write about, and it is the instrument you measure with. That dual role is why inventing questions is costly — a set built from what sounds plausible produces content nobody searches and measurements that cannot guide anything.
Treat the question set as a fixed asset. Changing it mid-programme invalidates every comparison you have made so far.
Step 1: Harvest what buyers already said
The best source is language that already exists in your business. Pull it in this order, because quality declines with each step.
- Recorded sales calls and support tickets — the actual phrasing, in context
- Your own site search queries, which show what people cannot find
- Questions in competitor reviews and community forums
- Search suggestions and related searches, used to find phrasings you missed
- Worst last: what the team assumes buyers ask. Assume wrong
Step 2: Normalise phrasing without losing intent
The same question gets asked many ways. Collapse those into one entry, but keep enough variants to match how people actually phrase it. The canonical form drives the content; the variants go into the measurement set so you test the way people ask, not the way you wrote it.
Keep the original wording in the record. Averaging five phrasings into one tidy question loses the signal that made the clustering useful.
Step 3: Sort by intent
Four buckets cover almost everything, and they behave differently enough to be worth separating.
- Informational: what is X, how does X work. Highest volume, lowest commercial intent
- Comparison: X vs Y, best X for Z. Where competitors appear, so it is highly competitive
- Transactional: X pricing, X free trial. Smallest volume, closest to revenue
- Navigational: brand plus product. Usually few, but they should already be won
Step 4: Find the gaps
Ask each question on three to five AI platforms and record whether you appear. Questions where a competitor appears and you do not are your priority list — that is the gap, measured rather than guessed.
Questions nobody's content addresses are the cheaper wins. There is no competition because nobody has written the answer yet.
- Competitor cited, you absent: highest priority
- Nobody cited: cheapest available position
- You already cited: maintain, and do not spend budget here
Step 5: Freeze it and reuse it
Lock the set, store it with timestamps, and use the same questions for every future measurement. When visibility changes six weeks from now, this set is what makes the change legible.
Review it quarterly at most. Adding questions is fine as long as the originals stay in place and you report on both.
A question set built from real conversations is reusable for years. One invented in a workshop is obsolete the moment it is finished. Sources reviewed on 2026-10-17. This article records a method with verifiable boundaries; it does not claim that any search or AI platform will index, cite or recommend anything.