To list prompts under an intent, use GET /intents/:intentId — the response includes a
prompts[] array.
Prompt styles
Every prompt has astyle tag. Variety matters: 5 paraphrases of “best meditation app” don’t tell you anything new, but a problem-led prompt and a comparison prompt for the same intent often produce very different recommendations from the same model.
The bootstrap aims to produce at least 3 different styles per intent. When you add prompts manually, picking a style outside what already exists for an intent gives you the most new signal.
Add a prompt
Body
Response
status: "active" and will be included in the next scan. If you want it picked up immediately, follow with POST /scan.
Deduplication. Re-posting the same text under the same intent is a no-op — we fingerprint the normalised text and ignore duplicates. This is safe to use idempotently from import scripts.
Edit / pause / archive
text re-fingerprints the prompt; if the new text matches an existing one under the same intent, the update will fail silently and you’ll get the existing row’s text on the next read.
Delete (archive)
Code examples
Tips for writing good prompts
- Keep it conversational. “best meditation app for beginners” beats “Comprehensive evaluation of meditation applications for novice users”.
- Avoid mentioning yourself. A prompt that names your app guarantees you’ll appear — but it doesn’t measure visibility, it measures the assistant’s willingness to repeat a name back.
- One question per prompt. Compound questions confuse the assistants and yield messier mention extraction.
- Cover the intent, not the feature. “an app that plays nature sounds” is too narrow if your intent is Reduce stress and anxiety to feel calm; “what helps me calm down before bed?” is better.
- Add at least one prompt in each language you care about — assistants behave differently across languages (a Turkish prompt often produces a different competitor set than an English one for the same intent).
Credits
Listing prompts is part of the intent drill-down (2 credits).
Errors
See also
- Intents — list/drill-down + intent CRUD
- Answers — see the actual AI responses for each prompt
- Scan & Settings — re-scan after adding new prompts

