---
name: geo-action-plan
description: >
  Generate a prioritized GEO action plan from Promptwatch visibility, content gap,
  citation, competitor, crawler, and AI visitor data. Use when someone wants the
  highest-impact actions to improve AI search visibility, citations, content coverage,
  crawler access, or AI-referred traffic.
---

# GEO Action Plan

## Prerequisites

Requires the Promptwatch MCP server. Add to your agent config:

```json
{
  "mcpServers": {
    "promptwatch": {
      "url": "https://server.promptwatch.com/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}
```

Get your API key at https://app.promptwatch.com -> Settings -> API Keys. Run `setup-promptwatch-mcp` if needed.

## Tools

- `list_projects`: resolve project.
- `get_visibility_time_series`: overall visibility trend.
- `get_content_gap_stats`: coverage distribution.
- `list_content_gap_prompts`: lowest-coverage prompts.
- `get_content_gap_recommendations`: per-prompt recommendations.
- `get_citations`: citation landscape.
- `get_citation_top_pages`: own pages that get cited.
- `get_self_citation_frequency`: self-citation trend.
- `get_competitor_heatmap`: competitor visibility by model.
- `get_top_competitors`: competitors with highest mention volume.
- `get_brand_visibility_over_time`: brand vs competitor trend.
- `get_crawler_trend`: AI crawler access trend.
- `get_top_crawler_pages`: pages crawlers visit most.
- `get_visitor_total`: AI-referred traffic volume.
- `get_visitor_referrer_stats`: AI traffic sources.

## Workflow

1. Resolve project.
2. Use a 30-day current period and matching previous period unless the user specifies a range.
3. Pull data across visibility, content gaps, citations, competitors, crawlers, and visitors. Parallelize when tools allow.
4. Identify confirmed problems only. Ignore weak signals with no data support.
5. Convert findings into max 10 prioritized actions.
6. Group actions by severity: `critical`, `high`, `medium`, `low`.
7. For action definitions and thresholds, use `references/action-types.md`.

## Prioritization Rules

- Lead with critical/high actions.
- Prefer high-impact, specific actions over broad advice.
- Back every action with a metric, prompt, page, domain, model, or competitor from Promptwatch.
- Do not invent recommendations when a dataset is empty.
- If many issues exist, keep the top 10 by likely GEO impact.

## Action Format

Each action must include:

- Title: short, concrete imperative.
- Severity: `critical`, `high`, `medium`, or `low`.
- Action type: `CONTENT_CREATION`, `CONTENT_OPTIMIZATION`, `CITATION`, `COMPETITIVE`, `TECHNICAL`, or `MONITORING`.
- Evidence: exact Promptwatch data point(s).
- Why it matters: GEO impact in one sentence.
- Next step: specific owner-ready task.
- Follow-up skill: `content-gap-audit`, `competitor-benchmark`, `citation-deep-dive`, `crawler-health-check`, or `ai-traffic-analysis`.

## Output

Produce:

1. One-paragraph health summary.
2. Prioritized action list grouped by severity.
3. Data caveats and empty datasets.
4. Suggested Promptwatch dashboard areas for deeper investigation.

If no meaningful data exists, output a setup checklist: tracked prompts, competitors, sitemap/crawler data, and visitor tracking.