seo-aeo-content-strategy

Files (2)

• SKILL.md
• content-brief-template.md · references/

SKILL.md (first lines)

## Why this matters

- **Topic clusters drive ~30% more organic traffic and hold rankings 2.5x longer** than standalone content (industry benchmarks, 2025).
- **44.2% of all LLM citations come from the first 30% of text** (large-sample ChatGPT citation analysis, ~1.2M pages). The intro carries disproportionate weight — front-load your answer.
- **78.4% of citations containing questions come from H2 headings** — LLMs treat H2s as prompts and the following paragraphs as answers.
- **Heavily-cited text has ~20.6% entity density** (brand names, tool names, proper nouns) vs. 5-8% for normal English.
- **53% of citations come from the middle of a paragraph**, not the first sentence. LLMs extract the sentence with the highest information gain, not the opener.
- **Content not updated quarterly is ~3x more likely to lose AI citations** (multi-model freshness benchmarks).
- **Brands are 6.5x more likely to be cited via third-party sources** than their own domains (vendor research, Oct 2025). Distribution matters as much as publication.
- **Earned media distribution increases AI citations up to 325%** vs publishing only on your own site (syndication research, Dec 2025).

All citations above are data-backed. The ski ramp framework has p-value 0.0 (statistically indisputable) in large-sample citation studies.

## Process

1. **Define business goals and audience.** What must the content produce — leads, trials, awareness, authority? Who specifically is the reader, in which intent state?
2. **Map the topic universe.** Core topics (3-7) → subtopics (5-15 per core) → supporting questions (unlimited). This becomes the cluster architecture.
3. **Design cluster architecture.** Pillar pages (broad, 2500+ words, link hub) → cluster pages (specific subtopics, link back to pillar + peers) → supporting content (FAQ, glossary, tools, comparisons).
4. **Audit existing content.** Score every page on the 1-5 matrix below: **keep**, **update**, **consolidate**, **prune**, or identify as a **gap** to create.
5. **Create content briefs with AEO structure baked in.** Use the ski ramp framework by default. See `references/content-brief-template.md` for the complete brief template.
6. **Build the editorial calendar.** Mix new creation with refresh cycles. Assume quarterly review of all priority content.
7. **Define measurement criteria.** What does success look like per content type? Rankings, traffic, AI citations, leads, revenue?

## Frameworks

### Topic cluster design template

```
Pillar page: "The Complete Guide to [Core Topic]"
├── Cluster page 1: "[Specific subtopic]"
├── Cluster page 2: "[Specific subtopic]"
├── Cluster page 3: "[Specific subtopic]"
│   ├── Supporting: FAQ
│   ├── Supporting: Comparison / vs
│   └── Supporting: Tool/calculator
└── Cluster page N: ...
```

**Linking rules:**
- Every cluster page links back to the pillar (upward)
- Pillar links to all clusters (downward)
- Clusters cross-link to 2-3 topically related clusters (sideways)
- Supporting content links to its parent cluster
- Anchor text is descriptive, not "click here"

**Sizing:** pillar pages 2,500-5,000 words; cluster pages 1,200-2,500 words; supporting content as needed. These are guides, not targets — write what the topic actually requires.

### Content audit scoring matrix

| Score | Label | Action | Criteria |
|-------|-------|--------|----------|
| **1** | Strong | **Keep / Monitor** | Ranking top 10, driving conversions, recently updated, AI-cited |
| **2** | Needs refresh | **Update** | Ranking 11-30, traffic declining, outdated info, structurally sound |
| **3** | Duplicative | **Consolidate** | Multiple thin pages on same topic, cannibalizing each other |
| **4** | Underperforming | **Prune or rewrite** | Not ranking, no traffic, no backlinks, thin content |
| **5** | Gap | **Create** | Topic covered by competitors but missing from our site |

**Pruning is not deleting.** A pruned page may be merged into another (301 redirect), rewritten entirely, noindexed (for legal/historical retention), or deleted with a redirect to the most relevant parent page.

### The "Ski Ramp" — AEO content structure

**This is the single most important content framework in this skill.** Derived from large-sample ChatGPT citation analysis (p-value 0.0). Apply it to every piece of content meant to earn AI citations.

**1. Front-load the answer (first 30% captures 44.2% of citations).**
- Put the key claim, definitive answer, or headline finding in the first paragraph.
- No throat-clearing intros. No "In this article, we'll explore…"
- Write the way a journalist writes a news lede: most important fact first.

**2. Use question-based H2 headings (78.4% of question-containing citations come from headings).**
- LLMs treat H2s as prompts and the following paragraphs as answers.
- Phrase H2s the way a user would ask the question: "How long does SEO take to show results?" not "SEO Timeline Overview."
- One topic per H2 section.

**3. Follow each H2 with a direct, definitive answer.**
- No hedging. "Teams that publish twice a week see 2.3x more traffic" beats "Publishing frequency may help traffic."
- The direct answer goes in the first 1-2 sentences after the heading. Supporting detail follows.
- LLMs penalize hedging language in the extraction pass.

**4. High entity density (~20%).**
- Use specific brand names, tool names, numbers, dates, proper nouns.
- "Increase budget by 15% in Q3" beats "increase budget in the second half of the year."
- Dates and numbers are universal citation positives — include real ones whenever possible.

**5. The middle of paragraphs matters.**
- 53% of citations come from mid-paragraph, not the opening sentence.
- Put the information gain — the specific, concrete, citable fact — somewhere in the middle of a 3-5 sentence paragraph, after context and before the transition.

**6. Include a summary/conclusion section (the "wake-up" zone).**
- The last 10% of content before the footer sees a citation bump.
- A short summary or "Key takeaways" block captures this.

**What NOT to do:**
- Don't artificially stuff entities or brand names — modern LLMs detect this and it damages trust signals.
- Don't build content just to be "summarizable." Large-sample work: making content summarizable alone doesn't make earning citations easier. Information gain and entity clarity do.
- Don't use clickbait H2s that don't match the question a user would actually ask. The AI matches on semantic intent, not clever copy.

### Content brief template (summary)

Core fields every brief should include:

| Field | Notes |
|-------|-------|
| Target keyword cluster | Primary + 5-15 supporting |
| Search intent | Informational / commercial / transactional / navigational |
| Fan-out sub-queries | From seo-keyword-research skill |
| Target word count | Based on top-10 SERP average + 20% |
| H1 | Primary topic, natural phrasing, includes target KW |
| H2 structure | **Question-based**, one topic each, sequenced by the user journey |
| First-paragraph key claim | The specific answer/finding to front-load |
| Key entities | Named brands, tools, people, places, dates, numbers to include |
| Internal links | To pillar (required) + 2-3 related clusters + supporting content |
| External citations | Primary sources, data, expert quotes |
| Schema type | Article, FAQPage, HowTo, Product, etc. |
| Competitor content to beat | Top 3-5 currently ranking |
| Information gain angle | What new data/insight do we add that nothing else has? |
| Named author + credentials | Required for E-E-A-T |

See `references/content-brief-template.md` for the complete version with prompts and guidance for each field.

From narayan-metaflow/metaflow-marketing-skills, MIT.

More from narayan-metaflow/metaflow-marketing-skills