Curate generated prompts
After PromptDNA generates your prompt set, curation keeps it lean, on-brand and aligned with how real users ask AI assistants about your category.
Open Settings β Sites β [your site] β Prompts to manage your active prompt list.
Approve
Newly generated prompts start in a draft state. Approve the ones that fit your brand, services and target audience. Only approved prompts are sent to AI providers during measurement cycles and counted toward your prompt quota.
Edit
You can edit a prompt's wording at any time. Common edits:
- Replace generic terms with the words your customers actually use.
- Remove brand names that bias the prompt.
- Adjust city wording (e.g. "in Zurich" vs "near Zurich").
Edits take effect on the next measurement cycle. Historical results for the previous wording are preserved.
Soft-delete
Deleting a prompt is a soft delete β the prompt is hidden from the active list and stops being measured, but all historical data (LLM responses, metrics, citations) is preserved.
Soft-deleted prompts:
- Do not count toward your prompt quota.
- Can be restored later without losing history.
- Are excluded from dashboards, reports and KPIs going forward.
Restore
Restore a soft-deleted prompt from the prompt archive (Settings β Sites β [your site] β Prompt archive). Restoring re-activates measurement on the next cycle and the prompt counts toward your quota again.
Identical prompt text can exist multiple times across different cities β the uniqueness rule is enforced per (prompt, city) combination, so multi-city prompts are not blocked by soft-deleted siblings.
Quality classification
Each active prompt is automatically classified as Healthy, Watch, Low or Excluded based on the last 15 LLM responses. See Prompt quality classification for what each label means and how to act on it.