SEO Article Writer

SEO Article Writer

Write articles that satisfy three audiences simultaneously: Google (ranking), AI models (citation), and humans (reading).

Workflow

Step 1 — Research (MANDATORY)

Before writing, gather SERP data. Two paths:

• No prior research: Invoke /seo-research [target keyword] first. Wait for the Research Brief output before proceeding.
• User provides research: Accept the existing brief and skip to Step 2.

Never write an article without a Research Brief.

Step 2 — Definition Engineering

The definition sentence is the single most important sentence in the article. It determines whether AI Overview, ChatGPT, and Perplexity will cite you.

The formula (derived from highest-citation articles: AWS, NVIDIA, Google ML Glossary):

[X] is a [category noun] that [verb1], [verb2], and [verb3] to [purpose] — [differentiating clause].

Rules:

• One sentence, 15-30 words
• Three-verb chain (perceive/understand + decide/reason + act/execute)
• End with "on behalf of [user]" or "to [achieve goal]" — highest AI citation rate
• No preamble. Definition IS the first sentence of the article (or paragraph 2 if using a hook opening)
• Include the primary keyword in its exact form

The definition paragraph (not just the sentence):

The definition sentence is the GEO citation anchor. But the definition paragraph does five jobs:

1. Definition sentence (15-30 words, X is a Y that [verb1], [verb2], [verb3])
2. Boundary line: What it's NOT, in one contrast. Name the adjacent category and the technical dividing line. ("Not a chatbot that generates text about your tasks, but an agent that executes them: books the meeting, sends the email, writes code when no pre-built tool exists.")
3. Modifier unpacking: If the term is [modifier] [noun] (e.g., "personal agent"), unpack the modifier with 2-3 concrete specifics. Don't just say the modifier matters — show what it adds.
4. Flywheel or mechanism: One sentence naming the core value loop. This is what makes the reader think "oh, that's why this category exists."
5. Verdict sentence (2-5 words): crystallize the whole paragraph. ("That flywheel is the product.")

The definition sentence is for AI citation. The definition paragraph is for human conviction. Both matter.

Disambiguation (if term has multiple meanings):

• Within first 100 words, acknowledge other meanings in one sentence
• Use parenthetical: "(In entertainment, a personal agent is a talent rep. This guide covers the technology definition.)"

Step 3 — Concept Precision Check (for technical categories)

When writing about a technology category (AI agents, cloud computing, blockchain, etc.), lock down the core terms before drafting. Imprecise terminology produces imprecise articles.

For each core term in the article:

1. Write a 1-sentence technical definition from first principles (e.g., "An agent is an LLM that can call tools and execute tasks in a loop, not just generate text")
2. Identify the boundary: What's the single thing that separates this term from its nearest neighbor? (Agent vs chatbot: tool calling. Personal agent vs AI agent: persistent user context.)
3. Lock the usage rule: Is the term a product name, a technical architecture, or a category label? These get used differently. ("Manus is a product. Agent is the engine behind it. Personal agent is the category.")

Run this check against every instance of the core terms in the draft. If a sentence uses the term in a way that contradicts the locked definition, fix it. This is especially critical for articles defining a category, where a single imprecise sentence can undermine the entire piece.

Step 4 — Article Structure

Structure is for Google and AI systems. It determines what gets indexed, extracted, and appears as rich results.

H1 (on-page heading):

What Is [X]? Definition, Examples, and How It Works [Year]

• Primary keyword in exact form
• Can be longer than 60 characters — this is the visible page heading, not the SERP title
• Year tag for freshness signal (on time-sensitive topics)

Title Psychology (derived from Every.to's highest-performing patterns):

Before optimizing for character count, nail the psychology of the title:

<title> tag (SERP title) — DIFFERENT from H1:

• Under 55-60 characters (Google truncates at ~580px)
• NO brand suffix — wastes 8-15 characters that should be keywords
• Primary keyword as close to beginning as possible

Meta description:

• 120-155 characters, complete sentence, includes primary keyword
• Functions as click-through pitch: "why should I read this over the 9 other results?"
• Formula: [Definition or core answer, ~80 chars] + [Value differentiator, ~60 chars]
• No brand name unless brand IS the topic. No quotes or special characters.

H2 headings — ALL as questions:

## What is [X]?
## Why does [X] matter?
## How does [X] work?
## What are the core capabilities of [X]?
## How is [X] different from [Y]?
## What can [X] actually do?
## Who is building [X] in [Year]?
## What are the challenges and limitations?
## How do you choose [X]?
## What is the future of [X]?
## FAQ

Question-format H2s align with highest-cited articles in AI Overview (AWS, NVIDIA, Google Cloud) and match "People Also Ask" queries directly.

Mandatory structural elements:

• Table of Contents with anchor links (before first H2)
• At least 2 comparison tables (against competitors/adjacent categories)
• At least 1 evaluation/criteria table
• Table humanization rule: Never drop a table cold. Every table needs a 1-2 sentence narrative intro that tells the reader what to look for and why it matters. The table delivers data; the intro delivers meaning.
• Comparison table precision rule: For each column in a comparison table, state both what the category CAN do and what it CANNOT do. "Answers questions in a conversation" is incomplete. "Answers questions in a conversation. No tool access, no task execution" draws a boundary. The absence is often more informative than the presence.
• Product list framing rule: Before any list of products, companies, or examples, establish the conceptual framework that organizes them. Never drop a bullet list cold. One sentence that teaches the reader the organizing principle. ("The products below are built on agents: LLMs that call tools, execute tasks, and operate in loops. The product is the interface. The agent is the engine.")
• FAQ section with 10+ Q&As
• "Challenges and Limitations" section (E-E-A-T signal — 7 of 8 top-ranking definition articles have this)
• Related Reading links (3-5 internal links at bottom)

Schema markup:

• Article (author, datePublished, dateModified, publisher, wordCount)
• BreadcrumbList
• Organization
• NOT FAQPage (restricted to government/healthcare since Aug 2023 — keep FAQ content for AI citation value)

Progressive Disclosure Principle:

Never bury the conclusion. Lead with the verdict. Let the article justify the opening claim.

Step 5 — Writing Style (Dan Shipper Method)

Load references/writing-style.md for the complete style guide.

Core principle: Write like a builder who just came back from the front line — with data, with scars, with thoughts about the bigger picture.

Key rules (always in memory — no need to load reference for these):

• First paragraph: 3 sentences max, personal micro-scene preferred
• Definition sentence goes in paragraph 2
• 30%+ of all paragraphs should be single-sentence paragraphs
• Every analytical passage ends with a verdict sentence (2-5 words)
• Conversational transitions only ("Here's what I mean." not "Furthermore")
• Max 2-3 em-dashes per 1000 words — overuse is the #1 AI writing tell
• At least 2 vivid verbs per section (not generic be/have/get/make/do)
• Person switching: narrative "I" → analytical "You/We" → conclusion "I"

Load references/writing-style.md for: the three opening moves, sentence rhythm targets, paragraph patterns, transitions list, voice toolkit, metaphor rules, section opening techniques, narrative arc, endings, and the full blacklist.

Step 6 — On-Page SEO

Keyword placement (primary keyword):

• Title tag: exact match
• H1: exact match or close variant
• URL slug: exact match, hyphenated
• First paragraph: within first 100 words
• Multiple H2 headings: natural inclusion
• Image alt text: at least 2 images
• Last paragraph / conclusion

Keyword density: 1-2% natural occurrence. Never force it.

Internal links: 3-5 per 1000 words. Anchor text uses primary or related keywords. Every article links to its topic cluster pillar page.

External links: 5-10 to authoritative sources (Wikipedia, official docs, research papers). Opens in new tab. Establishes E-E-A-T.

CTA strategy:

• 3-4 inline CTAs per article, spaced every 800-1200 words
• Placement: always AFTER a "problem → solution" passage, never random
• Progressive commitment: "Learn more" → "Try it" → "Sign up"
• All CTA URLs: ?from=blog-[slug]
• Category education phase: emphasize category word, not brand ("See how a personal agent handles this" > "Try [Brand] for free")

Open Graph tags: og:title, og:description (up to 200 chars), og:image, og:url

Step 7 — GEO (Generative Engine Optimization)

GEO determines whether AI models (ChatGPT, Claude, Perplexity, Google AI Overview) cite your article when answering user questions. SEO gets you ranked. GEO gets you quoted.

GEO citation priority (what AI models extract, in order):

Self-contained blocks:

• Every H2 section must be independently understandable and citable without context from other sections
• GEO atomic test: Take any section out of context. Can it be cited as a standalone answer? If not, add a framing sentence at the top.

Citable patterns (what AI models cite most):

1. Definition sentence in X is a Y that does Z format
2. Comparison tables with clear, search-intent-aligned column headers
3. Short declarative statements with specific facts/numbers
4. FAQ answers that start with the question's keywords
5. "Not X, but Y" contrast statements that draw category boundaries

FAQ as GEO weapon: FAQ answers are the second-highest citation target after the definition. Optimize each one:

• First sentence directly answers the question (zero preamble)
• Include the most recent verifiable data point (date, price, market size, product launch)
• Keep answers to 2-4 sentences for citation-friendly length
• Link to authoritative sources within answers for E-E-A-T

Step 8 — Meta Output Block (MANDATORY)

Every article ends with a <!-- SEO Meta --> block after Related Reading:

<!-- SEO Meta
meta_title: "[under 60 chars, no brand suffix]"
meta_description: "[120-155 chars, includes primary keyword, compelling sentence]"
h1: "[on-page heading, can be longer than title tag]"
og_title: "[same as meta_title or slightly adapted for social]"
og_description: "[same as meta_description or adapted, up to 200 chars]"
canonical_url: "https://ego.app/blog/[slug]"
primary_keyword: "[exact match keyword]"
secondary_keywords: "[comma-separated list]"
date_published: "[YYYY-MM-DD]"
date_modified: "[YYYY-MM-DD]"
-->

Hard rules:

1. meta_title ≠ h1 (title is SERP-optimized, H1 is page-optimized)
2. meta_title must NOT contain brand name, "| ego", "- ego Blog", or any suffix. Zero exceptions.
3. meta_description must be complete sentence, 120-155 characters, includes primary keyword.

Step 9 — Anti-AI-Slop Check

Load references/anti-ai-slop.md. Scan every draft against all 14 tells before delivery.

Quick-check (always in memory):

• Em-dashes: max 2-3 per 1000 words
• Zero banned AI vocabulary (delve, unpack, tapestry, leverage, holistic, robust, comprehensive, foster, harness, pivotal, embark, synergy, paradigm...)
• Contractions used naturally throughout
• Every claim has a specific name, number, or date
• At least 1 real failure/mistake/uncomfortable truth per article
• Read aloud test: must sound like a person, not a news anchor

Word Count Guidelines

Split rule: Any article (except pillar) exceeding 3,000 words with 2+ core arguments → split and interlink.

Education:Promotion Ratio

Article Type Templates

Pillar / Definition Article

[Hook paragraph — scenario the reader recognizes]
[Definition paragraph — formal definition + boundary-setting]

## Table of Contents
## Why does [X] matter?
## How does [X] work?
  [Scene opening → technical explanation wrapped in analogy]
  [Comparison table: your approach vs standard approach]
## What are the core capabilities?
  [Table: capability | meaning | without it you get...]
## How is [X] different from [adjacent categories]?
  [Full comparison table]
## What can [X] do? [Examples]
## Who is building [X]?
## Challenges and limitations
## How to choose
  [Evaluation table]
## Future
## FAQ (10-12 Q&As)
[Related Reading: 3-5 internal links]
<!-- SEO Meta ... -->

Comparison Article ("X vs Y")

[Hook — the confusion most people have]
[One-line thesis: the core difference]

## What is [X]?
## What is [Y]?
## [X] vs [Y]: key differences
  [Comparison table + subsections by dimension]
## When to use [X] vs [Y]
## FAQ
[Related Reading]
<!-- SEO Meta ... -->

Third-Road Article ("X vs Y: Why You Need Z")

[Hook — the false dilemma]

## What is [X]?
## What is [Y]?
## [X] vs [Y] compared [Fair, thorough comparison]
## The shared limitation
  [Both fail at the same thing — set up the pivot]
## A different approach: [Z]
  [Your category/product enters as the resolution]
## FAQ
[Related Reading]
<!-- SEO Meta ... -->

Checklist (Run Before Publishing)

Content Quality

• Hook in first paragraph — not a summary, not a definition-first opening
• Definition sentence follows X is a Y that [verb1], [verb2], [verb3] pattern
• Definition paragraph covers all 5 jobs: definition sentence, boundary line, modifier unpacking, flywheel, verdict
• Concept Precision Check passed (for technical categories): core terms locked with 1-sentence definition, boundary, and usage rule
• Disambiguation within first 100 words (if term has multiple meanings)
• All H2s are questions
• At least 2 comparison tables, each with narrative intro
• Comparison tables state both CAN and CANNOT for each column
• Product lists/examples prefaced with a conceptual framework sentence
• "Challenges and Limitations" section present
• FAQ with 10+ Q&As
• Each section passes GEO atomic test (citable independently)
• No "In today's rapidly evolving..." or similar AI-sounding phrases
• Short verdict sentences present (2-5 words after longer passages)

SEO

• Primary keyword in: title tag, H1, URL, first 100 words, 3+ H2s, alt text
• <title> tag under 60 characters, NO brand suffix, keyword front-loaded
• <title> tag is DIFFERENT from H1
• Meta description: 120-155 characters, complete sentence, includes primary keyword
• Meta description is NOT a repeat of the title tag
• TOC with anchor links
• 3-5 internal links per 1000 words
• 5-10 external links to authoritative sources
• 3-4 inline CTAs with tracking parameters
• Related Reading (3-5 links) at bottom
• datePublished and dateModified present
• Meta Output Block appended after Related Reading (Step 7)

Schema

• Article schema (author, dates, publisher, wordCount)
• BreadcrumbList schema
• Organization schema

Style (The Dan Shipper Test)

• First paragraph: 3 sentences max, uses personal micro-scene / poetic abstraction / provocative assertion
• 30%+ of paragraphs are single-sentence paragraphs
• Every analytical passage ends with a verdict sentence (2-5 words)
• At least 3 parenthetical asides with self-deprecating humor
• At most 2-3 em-dashes per 1000 words
• All data expressed as precise numbers — zero vague adjectives
• At least 3 analogies from everyday domains (food, body, school, relationships)
• Technical concepts: everyday analogy → parenthetical definition → immediate example
• At least 2 vivid verbs per section (not generic be/have/get/make/do)
• Zero academic transition words (Furthermore/Moreover/In addition/Additionally)
• Zero "You should..." sentences
• Ending: circles back to opening scene or lands a quotable gold sentence
• Person switching: narrative "I" → analytical "You/We" → conclusion "I"

Anti-AI-Slop (run after every draft)

• Em-dash count: max 2-3 per 1000 words in prose only (exclude tables, bullet labels, meta lines)
• Zero banned AI vocabulary (delve, unpack, multifaceted, myriad, tapestry, leverage, utilize, holistic, robust, comprehensive, foster, facilitate, harness, underscore, pivotal, paramount, embark, realm, intricate, cutting-edge, groundbreaking, navigate, synergy, paradigm)
• Contractions used naturally throughout ("it's" not "it is", "don't" not "do not")
• Every claim has a specific name, number, or date — no confident generalities
• Max 1 "It's not about X, it's about Y" construction per article
• Lists/bullets only for genuinely parallel items — narrative ideas stay in prose
• Every section has clear point of view / verdict — not a neutral overview
• Zero emojis
• Zero hedge stacks ("may potentially help to possibly...")
• At least 1 real failure/mistake/uncomfortable truth per article
• Sentence length varies wildly (3-word punches mixed with 30-word explorations)
• Symmetric triplet check: no more than 3 "X, Y, and Z" three-element lists in the article
• Read the draft aloud — if it sounds like a news anchor, rewrite