> ## Documentation Index
> Fetch the complete documentation index at: https://docs.frictionai.co/llms.txt
> Use this file to discover all available pages before exploring further.
# Actions: Prompts
> Find underperforming prompts and inspect the response evidence behind recommended fixes.
# Actions: Prompts
Actions: Prompts surfaces prompts where the brand underperforms across visibility, sentiment, or purchase intent. It is the fastest way to move from score diagnosis to a fix list.
## What appears here
Prompts where the brand is weak, missing, negatively framed, or losing the recommendation.
Filter by all metrics or a single metric: visibility, sentiment, or purchase intent.
Filter by provider and inspect how each model responded.
Open prompt evidence, response text, recommendations, sources, and search keywords.
## Filters
Use all metrics for triage, then isolate visibility, sentiment, or purchase intent when you need a cleaner fix list.
Filter by custom prompts or system prompts so you know whether the issue came from your own tracking set or brand-health analysis.
Filter to one provider when a weak answer appears to be model-specific.
Search prompt text when you are investigating a category, product line, campaign, or competitor.
## Use the detail drawer
Click an underperforming row to open the right-side detail panel.
Confirm which metric is weak and how severe the issue is.
Look at the explanation, response text, competitor mentions, source evidence, and recommendation cards.
Most fixes are content coverage, source quality, comparison positioning, product proof, or entity clarity.
## Fix patterns
Create or improve content that directly answers the missing prompt intent. Strengthen source authority and category association.
Address the source or narrative creating cautious, mixed, or negative framing. Add recent proof, reviews, customer outcomes, or corrections.
Improve comparison, use-case, pricing, review, and product proof pages so AI has evidence to recommend you.
Prioritize prompt blindspots that repeat across providers or map to high-value commercial intent. One weak answer is less important than a repeated pattern.