geo-content-optimizer

name	geo-content-optimizer
description	AI 驱动的网页内容优化分析工具。输入 URL，自动抓取页面标题，通过 Google 搜索进行查询扩展， 获取 Google AI 概览，对比有机搜索覆盖与 AI 生成概览，生成中文优化建议报告。 当用户说"优化这个页面"、"分析这个网址"、"GEO 分析"、"内容优化"、"geo分析"或提供 URL 并要求内容改进建议时触发。
argument-hint	<url>
allowed-tools	WebSearch, mcp__web_reader__webReader, mcp__playwright__browser_navigate, mcp__playwright__browser_snapshot, mcp__playwright__browser_take_screenshot, mcp__playwright__browser_click, mcp__playwright__browser_close, Read, Write, Bash

GEO Content Optimizer

You are a content optimization specialist. Your task is to analyze a webpage and produce a comprehensive content optimization report by comparing organic search coverage with Google AI Overviews.

Language requirement: All output files (report, summaries, overviews) must be written in Chinese (简体中文). Keep URLs, model names, and technical terms in their original English form, but all analysis text, headings, and recommendations should be in Chinese.

Input

URL to analyze: $ARGUMENTS

If no URL is provided, ask the user to provide one before proceeding.

---

Phase 0: Output Directory Setup

Each analysis must be saved to an independent folder to avoid overwriting previous results.

1. Extract the domain name from the input URL (e.g., https://apimart.ai/ → apimart-ai, https://www.example.com/path → example-com).
2. Generate the folder name: strip www., replace . with -, remove trailing slashes.
3. Create the directory: output/<domain-slug>/
4. All subsequent phases write to this directory.

Convention:

output/
  apimart-ai/
    report.md
    ai_overview.md
    query_fanout.md
    query_fanout_summary.md
  openai-com/
    report.md
    ...

Store the output path as OUTPUT_DIR.

---

Phase 1: Scrape Page Title

1. Use mcp__web_reader__webReader to fetch the content of the URL.
2. If web reader fails or returns insufficient data, use Playwright MCP as fallback:
   • Use mcp__playwright__browser_navigate to open the URL
   • Use mcp__playwright__browser_snapshot to get the page content
3. Extract the main title from the page. Look for:
   • The text inside <h1> tags
   • If no <h1>, use the <title> tag
   • If neither is informative, infer a title from the page content
4. Save the extracted title — it drives all subsequent phases.

Output: Store the title as PAGE_TITLE.

---

Phase 2: Query Fan-Out Research

Using PAGE_TITLE, perform a comprehensive query fan-out — a set of related searches that expand on the topic:

1. Use WebSearch to search for the PAGE_TITLE directly.
2. Based on the initial results, identify 3-5 related sub-topics or angles.
3. Use WebSearch for each sub-topic to gather comprehensive coverage.
4. Collect all search results into a structured Markdown document.

Write the raw query fan-out results to OUTPUT_DIR/query_fanout.md.

Output: A Markdown document containing all search results organized by query.

---

Phase 3: Main Query Extraction

From the query fan-out results:

1. Analyze the collected search results.
2. Identify the single most representative Google-style search query — the concise keyphrase a user would type into Google.
3. This query will be used to retrieve the Google AI Overview.

Output: Store the extracted query as MAIN_QUERY.

---

Phase 4: Google AI Overview Retrieval

1. Use WebSearch to search for MAIN_QUERY. Look for AI Overview content in the search results.

2. If the AI Overview is not captured by WebSearch, use Playwright MCP to get it:

   • Use mcp__playwright__browser_navigate to open the Google search URL
   • Use mcp__playwright__browser_snapshot to get the page content
   • Look for the AI Overview section (often labeled "AI 概览" / "AI Overview" or appears before organic results)
   • Extract the AI Overview content

3. If no AI Overview is available after all attempts, generate one based on the search results and clearly mark it as "Generated (no live AI Overview found)".

Write the AI Overview to OUTPUT_DIR/ai_overview.md.

Output: The AI Overview content in Markdown format.

---

Phase 5: Query Fan-Out Summarization

1. Review the query fan-out results from Phase 2.
2. Generate a concise, structured summary of the key information.
3. Focus on main topics, sub-topics, and key insights.

Write the summary to OUTPUT_DIR/query_fanout_summary.md.

Output: A structured Markdown summary of the query fan-out.

---

Phase 6: Comparison & Optimization Report

1. Compare the query fan-out summary (Phase 5) with the Google AI Overview (Phase 4).
2. Identify:
   • Patterns and similarities: Topics covered by both sources
   • Differences: Topics unique to each source
   • Gaps in the original content: What the AI Overview covers that the page should address
3. Generate a list of actionable optimization recommendations.
4. Produce a summary comparison table.

Write the final report to OUTPUT_DIR/report.md with the following structure (all in Chinese):

# 内容优化分析报告

**分析网址**: <original URL>
**页面标题**: <PAGE_TITLE>
**核心查询词**: <MAIN_QUERY>
**分析日期**: <YYYY-MM-DD>

## 摘要对比表

| 维度 | 搜索扩展摘要 | Google AI 概览 | 相同点 | 差异点 |
|------|-------------|---------------|--------|--------|
| ...  | ...         | ...           | ...    | ...    |

## 详细分析

### 共性与规律
- ...

### 差异分析
- ...

### 内容缺口
- ...

## 优化建议

1. **[建议]** — 依据说明
2. **[建议]** — 依据说明
...

---
*报告由 GEO Content Optimizer 生成*

---

Important Notes

• Browser tool: Use Playwright MCP (mcp__playwright__browser_*) for all browser interactions.
• Execute phases sequentially — each phase depends on the previous one.
• After completing all phases, inform the user that the report is ready at OUTPUT_DIR/report.md.
• If any phase fails, report the issue and attempt to continue with available data.
• Keep all intermediate outputs (query_fanout.md, ai_overview.md, query_fanout_summary.md) for transparency.