---
name: content-optimizer
description: Iterate a draft against the SEO AI Regent scoring engine until it clears the 70-point Review Gate, improving Content Score and GEO Score signal by signal. Use when asked to "optimize this draft", "score my article", or "make this pass the review gate".
---

# Content Optimizer

You are optimizing draft content with the `seo-ai-regent` MCP server's
`score_content` tool. The Content Score must reach **70/100** to pass the
Review Gate; the GEO Score measures AI-search citability.

## Workflow

1. Confirm the target keyword. If the user did not give one, ask - scoring
   is keyword-relative.
2. Call `score_content` with the keyword and the current draft.
3. Read the signal breakdown and improve the **lowest-weighted-down signals
   first**:
   - `termFrequency`: use the keyword and close variants naturally in the
     title, first paragraph, and headings.
   - `headingStructure`: one H1, logical H2/H3 hierarchy, question-style
     H2s help GEO's answerFormat too.
   - `wordCount`: substantive coverage, not padding.
   - `readability`: 10-22 words per sentence on average.
   - `internalLinks`: add relative links (`/path`) to related pages.
   - GEO `factualDensity` and `answerFormat`: lead sections with direct,
     self-contained answers; include concrete facts, numbers, and sources.
4. Apply edits to the draft, then call `score_content` again. Repeat until
   the Content Score is >= 70, or two consecutive iterations stop improving
   the score - then tell the user what is holding it back and why.
5. Report before/after scores and the signal deltas, so the user sees
   exactly what each edit bought.

## Rules

- Edit the user's draft; do not replace it with generated boilerplate.
- Never claim the content will rank or be cited - the score measures
  readiness, not outcomes.
- Preserve the author's voice and factual claims; flag, don't invent.