A prompt set is not a keyword list

The unit of measurement changed

In search, the unit is a query fragment — two or three words, stripped of intent, matched against a corpus. In AI answers, the unit is a question with a decision behind it: which tool should I use, is this vendor credible, what happens if I do this instead.

If you export your keyword list and paste it in as prompts, you will measure something, but it will not be the thing that matters. "aeo tools" as a prompt produces a generic listing answer. "Which AEO tool should a five-person marketing team pick if they cannot afford Profound?" produces the answer a real buyer receives — and the two have different sources, different competitors named, and different implications for what you should do next.

The four kinds of prompt worth tracking

A prompt set that only asks one kind of question gives you a one-dimensional picture. Cover the decision journey instead.

The recommendation prompts are the ones that hurt when you lose them, because they are the moment the engine picks someone. The accuracy prompts are the ones that hurt when you win them wrongly — being present in an answer that misdescribes you is worse than being absent.

• Category — what is this, how does it work, what should I know. You are competing with explainers here, not vendors.
• Comparison — X versus Y, alternatives to X. The competitive set becomes visible in these.
• Recommendation — what should I use, best tool for a situation. This is where the purchase decision happens.
• Brand — what is your company, is it any good, what does it cost. This is where accuracy problems surface.

Ten is a real constraint, not a starter allowance

Ten prompts is what the free tier tracks, and it is genuinely tight. But a focused ten beats a sprawling hundred for learning, because with ten you will actually read every answer, and reading the answers is where the insight is.

Teams with large prompt sets frequently stop reading the responses and start watching only the aggregate score. That is how a programme becomes a dashboard that nobody acts on. Start small enough that you read everything.

Where this lives in SolCrys

• Prompts live on the workspace prompts surface, where you can add them manually or generate a starter set.
• The generator proposes prompts from your brand and category — useful as raw material, but edit them. A generated set is a starting point, not a decision.
• Disabling a prompt frees its slot, so a mistake is recoverable without wasting the allowance.

Check yourself

• Take one keyword from your existing SEO list and rewrite it as a decision a buyer is trying to make.
• Which of the four prompt types is missing from the set you were planning?
• Why is a recommendation prompt more commercially loaded than a category prompt?

Go deeper

This lesson is the spine. These guides are the depth.

Continue

• Previous: Brand facts and the competitive set
• Next: Choosing your ten prompts