“We want to measure whether AI recommends us. Where do the questions come from?” The prompt bank is the instrument. Questions shape what the study can measure, so document how themes and prompts were selected. This is how a bank gets built so the numbers mean something.
Start from themes, not from the brand
Buyers often ask about a situation within a category rather than naming the category alone. Start by listing themes supported by documented demand or customer language. Possible inputs include:
- Search Console queries, if you have access
- Google autocomplete and People Also Ask for the category, where available
- Forums and communities where the category gets discussed
- Review platforms, for the language of satisfaction and complaint
- Call center FAQs and internal site search
A sunscreen category, used here as a worked example, yields themes like: protecting children, sensitive skin, daily facial use, beach and sport, ingredient safety, price and value, and where to buy. These are illustrative themes; validate them against available customer and search evidence for the category.
Give every theme several real angles
One question per theme is a sample of one, and assistants answer different angles differently. Four angle families, worked on the same theme:
- Who is asking: “sunscreen for a baby under six months old”
- The situation: “how do I keep my children from getting sunburned at the beach”
- How far along the decision is: “which sunscreen for children is best”
- The comparison being weighed: “mineral or chemical sunscreen for children”
An assistant answers those four in four different ways. They are not one question reworded, and the bank needs the spread.
Three layers, reported separately
The bank holds three layers, and each answers a different question:
- Unbranded. Questions that never name you. These measure spontaneous presence in that set of questions; they do not replace direct questions about the brand.
- Brand-named. What happens when someone names you: what the model believes, which critique arrives attached, where it sends buyers, whether your subbrands get credited.
- Competitive. Who owns the conversation: head-to-heads, and the comparisons happening without you in them.
Our method reports each layer separately because each measures a different outcome. We do not average the layers into one figure. There is no universal question count that makes every layer reportable: set the bank size from the agreed scope, questions, models, repetitions and reporting needs, and show the denominator for each result.
The three failure modes of homemade banks
1. The questions name the brand. Brand-named questions measure how the assistant responds when asked directly. They cannot by themselves measure spontaneous presence, which is why the unbranded layer has a separate role.
2. One question per theme. A single phrasing may not cover the different ways buyers approach a topic. Include relevant angle variety, then report how the questions were selected.
3. The wrong competitor set. A competitor set based only on internal assumptions may omit alternatives buyers compare. Use category research and observed answers to define and review the set; do not assume the answer set matches a sales team’s list.
Each failure corrupts a different number, silently. That is what makes banks worth building carefully.
The frozen-bank rule
Once the bank produces its first baseline, it freezes. New questions added later start their own baseline, and banks are never spliced into a trend. A later run asks the same questions so observed rates can be compared. The comparison shows what changed, not what caused the change. Document the bank so any future edit is deliberate, not accidental.
DIY or done with you
Everything above is buildable in-house with research tools and discipline. If you want it built or built together, that is the service, and it is the same list you just read applied to your category: themes with documented demand, angles validated against real phrasing, per-question origin annotations, and a coverage report: /banco-de-prompts/.