42 Seo Agi

Research Scripts

Use $SKILL_ROOT in all script calls:

# Full competitive research (SERP + keywords + competitor content analysis)
python3 "${SKILL_ROOT}/scripts/research.py" "<keyword>" --output=brief

# Detailed JSON output for deep analysis
python3 "${SKILL_ROOT}/scripts/research.py" "<keyword>" --output=json

# Google Search Console data (if creds available)
python3 "${SKILL_ROOT}/scripts/gsc_pull.py" "<site_url>" --keyword="<keyword>"

# Cannibalization detection
python3 "${SKILL_ROOT}/scripts/gsc_pull.py" "<site_url>" --keyword="<keyword>" --cannibalization

# Mock mode for testing (no API keys needed)
python3 "${SKILL_ROOT}/scripts/research.py" "<keyword>" --mock --output=compact

IMPORTANT: Always combine the skill root discovery and the script call into a single bash command block so the variable is available.

API Key Configuration

Keys are loaded from ~/.config/seo-agi/.env or environment variables:

DATAFORSEO_LOGIN=your_login
DATAFORSEO_PASSWORD=your_password
GSC_SERVICE_ACCOUNT_PATH=/path/to/service-account.json

MCP Tool Integration

If the user has Ahrefs or SEMRush MCP servers connected, use them to supplement or replace DataForSEO:

• Ahrefs MCP: site-explorer-organic-keywords, site-explorer-metrics, keywords-explorer-overview, keywords-explorer-related-terms, serp-overview for keyword data, SERP data, competitor metrics
• SEMRush MCP: keyword_research, organic_research, backlink_research for keyword data, domain analytics
• Use DataForSEO for content parsing (competitor page structure, headings, word counts) which MCP tools don't cover
• When multiple sources are available, cross-reference for higher confidence

Data Cascade (use in order of availability)

What the Research Gives You

The research script outputs:

• SERP data: Top 10 organic results with URLs, titles, descriptions
• Competitor content: Word counts, heading structures (H1/H2/H3), topics covered
• Related keywords: With search volume and difficulty scores
• PAA questions: People Also Ask questions for FAQ sections
• Analysis: Search intent detection, word count stats (min/max/median/recommended range), topic frequency across competitors, heading patterns

Use this data to inform every decision: word count targets, heading structure, topics to cover, questions to answer, competitive gaps to exploit.

1. CORE BELIEF SYSTEM

1. AI content is not the problem; generic content is. Do not rewrite the first page of Google. Add genuinely useful, sourced, less-common information.
2. Write for LLM Retrieval. The page must be easy to extract, summarize, cite, and quote by both search engines and AI answer engines.
3. Entity Consensus over Backlinks. LLMs trust brands mentioned consistently across high-signal domains (Reddit, Wikipedia, LinkedIn, Medium). Build consensus across platforms, not just link equity.
4. Tables are Mandatory. Use clean HTML <table> elements for cost, comparison, specs, and local services. Never simulate tables with bullet points.
5. Top-of-Page Dominance. The most important, answer-forward material goes at the absolute top. A fast-scan summary block must appear within the first 200 words.
6. Brand > Links. Google and LLMs prioritize "Brand + Keyword" searches. If ChatGPT doesn't know a website exists, a guest post there is worthless for GEO.

2. GOOGLE AI SEARCH -- 7 RANKING SIGNALS

Every piece of content is scored against these seven signals in Google's AI pipeline. Optimize for all seven.

3. THE 500-TOKEN CHUNK ARCHITECTURE

Google's AI retrieves content in ~500-token (~375 word) chunks. LLMs chunk at ~600 words with ~300 word overlap. Structure every page to feed this pipeline perfectly.

Chunk Rules:

• Question-Based H2s: Every H2 must match a real search query or a "Query Fan-Out" question (the logical follow-up an AI will suggest). Use PAA data from research to inform these.
• The Snippet Answer: The first 2-3 sentences immediately following any H2 must be a direct, concrete answer to that heading. No preamble. No definitions.
• The Contrast Statement: Within the chunk, include explicit X vs. Y comparisons with numbers (e.g., "Economy lots cost $16/day but require a 15-minute bus ride; terminal garages cost $43/day with direct skybridge access").
• Self-Contained Chunks: Never split a data table across chunk boundaries. Never stack two H2s without at least 250 words of substantive data between them.
• Front-Load Strength: The strongest content (bottom line, key recommendations) must appear in the first 3 chunks, not the last. AI retrieval may never reach buried material.

4. SEAT SIGNALS (Semantic + E-E-A-T + Entity/Knowledge Graph)

Semantic Keywords

Every page must cover:

• Primary head terms (from research: target keyword)
• Semantic neighbors (from research: related keywords and topic frequency data)
• Geo-modifiers (neighborhoods, nearby cities, landmarks served)
• Mode competitors (transit, taxi, Uber/Lyft, rideshare -- must be named even if you don't sell them)
• Operational terms (from research: common heading topics across competitors)

E-E-A-T Signals

• Experience: Location-specific operational details (terminal pickup spots, timing, traffic)
• Expertise: Pricing comparisons with real numbers, not vague "affordable" language
• Authority: Cite official sources (airport authority, transit authority, published fare schedules)
• Trust: Honest "Not For You" sections, transparent comparison against non-parking options

Entity / Knowledge Graph

Google's KG uses different NLP than transformers. Entity signals must be explicit:

• Full official entity names at least once (e.g., "Hartsfield-Jackson Atlanta International Airport" not just "ATL")
• Terminal numbers/names as distinct entities
• Airline-to-terminal mappings where relevant
• Parking lot names as entities, not just list items
• Operating authority names (Port Authority, airport authority, etc.)

5. QUALITY & AUDIT FILTERS

Before completing any output, pass these tests. If the content fails, rewrite it.

A. The Reddit Test

If this page were posted to a relevant subreddit, would a knowledgeable practitioner call it "AI slop" or ask "Where is the real data?"

Passing requires at least three of the following:

1. A hard number from an official or overlooked source (capacity, square footage, wait time, frequency, volume)
2. A layout or navigation detail only someone familiar with the place would know
3. A cost comparison that does real math (e.g., "5 days at $20/day = $100; an Uber round trip from downtown is roughly $30 total -- the break-even is about 2 days")
4. A schedule or operational detail with specifics (shuttle runs every X minutes; lot fills by Y time on Z days)
5. A "the thing they moved / changed / broke" detail -- something that changed recently
6. A real gotcha or failure mode described with enough specificity that a reader thinks "that happened to me"

B. The Prove-It Details

At least two hard operational facts must be present in every document:

• Capacity, frequency, fill rate, wait time, or distance measurements
• Break-even cost math showing when one option beats another
• Layout/navigation details that help someone who has never been there
• A recent change not yet reflected on most competing pages

C. The "Not For You" Block

Every page must include a section honestly telling the reader when this option is a bad fit. Name the specific scenario. Include at least one line a competitor would never say because it might scare off a lead. This is the ultimate E-E-A-T trust signal.

D. The Information Gain Test

A page passes when it contains content that cannot be found by reading the top 10 Google results for the same query. Use the research data to identify what competitors cover, then find what they miss.

6. TECHNICAL MARKUP RULES

The RDFa Hack

LLMs often ignore JSON-LD in the header. Embed semantic data directly inline using RDFa or Microdata (<span> tags). This is "alt-text for your text" -- label entities, costs, and services explicitly within paragraph code so LLMs extract it effortlessly.

Required Schema Per Page Type:

• FAQPage: Wrap every question-based H2 + answer pair
• HowTo: Any step-by-step booking or pickup process
• Product/Offer: Pricing tables and service options
• LocalBusiness: For facilities or lots listed
• BreadcrumbList: Site navigation context

See references/schema-patterns.md in the skill root for JSON-LD templates. Read it with: cat "${SKILL_ROOT}/references/schema-patterns.md"

Schema Serves 3 Independent Functions:

7. VERIFICATION & TAGGING SYSTEM

You are forbidden from inventing fake studies, statistics, or pricing. Use auditable tags for human editors.

Source Citation Rules:

Do not cite vaguely. Never write "official airport website" or "government data."

Instead cite specifically:

• "Broward County Aviation Department -- FLL Parking Rates (broward.org/airport/parking)"
• "FLL Airport Master Plan, 2024 update, Section 4.2"
• "FDOT Traffic Count Station 0934, I-595 at US-1 interchange"

8. REQUIRED PAGE STRUCTURE

Use this structure unless the brief explicitly requires something else.

1. Title

Clear, includes the main topic naturally, not overstuffed, promises a concrete outcome.

2. Opening Answer Block (first 100-150 words)

Answer the main query directly. Explain what makes this page useful or different. Preview the most important distinctions.

3. Fast-Scan Summary (immediately after opening)

One of: bullet summary (3-5 bullets max, each with a concrete fact), key takeaways box, comparison table, or quick decision matrix. Not optional. Every page needs a scannable extraction target near the top.

4. Main Body with Distinct Sections

Every section must do one unique job: explain, compare, quantify, define, rank, warn, price, or instruct. No filler sections. Use research data to determine which sections competitors cover and where the gaps are.

5. Comparison Table

Real HTML <table> with columns that do real work. Prefer: "Best For" (who should choose), "Main Tradeoff" (what you give up), "Why It Matters" (implication, not just fact), "Typical Cost" with {{VERIFY}} tags.

6. Prove-It Section (Information Gain)

The material that passes the Reddit Test. At minimum two hard operational facts with traceable citations.

7. Not For You Block

Specific scenarios where this is the wrong choice. At least one line a competitor would never publish.

8. Conclusion / Next Step

Direct. Summarize the decision and next action. Do not restate the entire page.

9. ABSOLUTE WRITING RULES

Never Do:

• Generic intros or definitional preambles
• "In today's fast-paced world" or any variant
• "Whether you're a ... or a ..." constructions
• The word "nestled"
• Em dashes
• Repetitive FAQ fluff
• Bulleted lists pretending to be tables
• Near-identical sections with only wording changes
• Empty headings without content
• Generic praise repeated across all items in a listicle
• Keyword stuffing
• Jump-link TOC patterns that create weak fragment URLs
• Content that sits outside your core service topical circle (a wildlife recovery site does not need a post on the industrial uses of guano — wide topical circles dilute AI authority signals and confuse intent classification)

Always Do:

• Short to medium sentences, concrete nouns, explicit comparisons
• Numbers and specifics over adjectives
• Entity-rich language (real product names, locations, service names)
• Honest negative recommendations alongside positive ones
• Front-load the strongest material

10. VERTICAL-SPECIFIC INSTRUCTIONS

Airport / Parking / Transportation Pages

1. Terminal-to-facility map or guide. List which airlines operate from which terminals and which parking option serves each best.
2. Capacity or availability context. How many spaces? When does it fill? What happens when full?
3. Rideshare/transit comparison math. Break-even calculation: at how many days does parking cost more than two Uber rides?
4. Pickup/dropoff operational details. Where exactly is rideshare pickup? Cell phone lot? What confuses first-timers?
5. Shuttle details. Frequency, hours, known reliability issues.
6. Peak-day warning. Name specific days or events that cause fill-ups. Not "busy periods" -- "cruise ship Saturdays," "Thanksgiving Wednesday."

Local Service Pages

• City/area naturally in title and opening
• Cost or pricing expectations with ranges
• Practical comparison table (service type vs. cost, emergency vs. standard, residential vs. commercial)
• Buyer questions people actually ask

Ask Maps & Conversational GBP Optimization

Google Maps and similar platforms are rolling out "Ask Maps" features — natural language queries like "who is open this Sunday?" or "who has same-day availability in [City]?" The answer is pulled from structured GBP data, not from your website.

Required data points to answer conversational queries:

• Hours with holiday/exception hours explicitly set
• Services listed as discrete GBP service items (not just in description prose)
• Q&A section pre-populated with the exact questions customers ask
• Posts updated at least bi-weekly (freshness signal for conversational pull)

Rule: If your GBP cannot answer "who has [service] available [specific condition]?" in structured form, a competitor with complete data wins that query even if your organic rankings are higher. Treat GBP structured fields as AEO markup, not optional admin work.

Listicles

• Each item must be substantively different
• Format per item: name, who it's best for, why it made the list, one differentiator, one tradeoff
• Strongest items first. Do not pad to reach a number.
• Segment large lists into logical groups (best budget, best premium, best for beginners)

Comparison / Pricing Pages

• Comparison table high on the page
• Clear selection criteria and who each option is best for
• Real tradeoffs, not just feature checkmarks
• Simple recommendation logic

11. LLM / AEO CITATION STRATEGY

LLMs pull from positions 51-100, not just page 1. Being the most structured and honest comparison page can earn AI citations even without traditional page 1 rankings.

To become citable by AI answer engines:

• Have the most complete pricing comparison (all modes, all price points)
• Have the clearest structured data (tables with labeled columns)
• Have unique operational content (terminal pickup, process steps)
• Be the page that tells the truth when competitors don't

Off-Page Sequencing — Existence Before On-Page

Google and AI agents now cross-check third-party signals before trusting your own site or Google Business Profile (GBP). An "inspector" layer verifies external mentions to filter spam. If the business doesn't exist in the wider web, on-page SEO and GBP submissions underperform or fail verification.

Required sequence:

1. Establish brand footprint first: Facebook page, industry-specific citation sites, press mentions, or PR outlets
2. Then submit or optimize the GBP
3. Then build on-page content — now the AI has external corroboration to amplify rather than question

Skipping step 1 is the most common reason a legitimate local business struggles to rank despite having a clean, well-structured site.

Entity Consensus Generation:

When prompted for broader strategy, output variations of core 500-token chunks formatted for cross-posting on LinkedIn, Medium, Reddit, and Vocal Media to build brand authority where LLMs scrape.

Reddit — Subdomains Over Standard Posts

Reddit is pulled into AI Overviews and conversational search results at high frequency, but standard www.reddit.com posts are often flagged as spam before indexing. Reddit operates dozens of subdomains treated by Google as distinct entities.

Tactical note: When seeding Reddit for entity consensus, explore indexed subdomain entry points beyond the standard www. Content indexed across multiple Reddit layers increases the probability of being retrieved in "Ask"-style conversational queries. Monitor which subdomain posts get crawled via Google Search Console and prioritize those paths for future brand mentions.

RAG Targeting — Write for AI Retrieval, Not Keyword Volume

Modern AI search agents (Gemini, ChatGPT, Perplexity) use Retrieval-Augmented Generation (RAG): they pull the most authoritative chunk available and surface it as the answer. This means zero-volume long-tail queries matter.

How to execute:

• Identify esoteric, service-specific questions your clients actually ask in sales calls or support tickets — even if keyword tools show "0 searches/month"
• Write a dedicated 500-token chunk answering each question with hard specifics
• These chunks "train" AI models to associate your domain with that competency, making you the cited source when a user asks the same question inside a chat interface

Rule: At least 20% of a content calendar should target zero-volume long-tail queries that demonstrate deep operational expertise. Traffic is a lagging indicator; AI citation is the leading one.

12. HUB & SPOKE INTERNAL LINKING

• Hub page = main topic page (e.g., "ATL Airport Parking")
• Spoke pages = detail pages, hotel pages, destination pages, supplier profiles, terminal guides
• Every spoke links back to its hub
• Hub links to its most important spokes
• Dead-end content (flat lists with no links) wastes crawl equity
• Use research data to identify which hub/spoke pages competitors link between

13. EXECUTION PROTOCOL

When the user provides a target keyword and brief:

1. Research: Run the data layer (combine discovery + script in one bash block):

   for dir in "." "${CLAUDE_PLUGIN_ROOT:-}" "$HOME/.claude/skills/seo-agi" "$HOME/.agents/skills/seo-agi" "$HOME/.codex/skills/seo-agi" "$HOME/seo-agi"; do [ -n "$dir" ] && [ -f "$dir/scripts/research.py" ] && SKILL_ROOT="$dir" && break; done; python3 "${SKILL_ROOT}/scripts/research.py" "<keyword>" --output=json
   

   If the script exits with an error (no DataForSEO creds), fall back in this order:

   • Try Ahrefs MCP tools (serp-overview, keywords-explorer-overview) if available
   • Try SEMRush MCP tools (keyword_research, organic_research) if available
   • Use WebSearch tool as last resort to manually research the SERP landscape Also search for official source pages, operational documents, recent changes, layout details, comparable cost math, and community feedback.

2. Brief: If the user did not provide a brief, build one:

   Topic: [inferred from keyword]
   Primary Keyword: [target keyword]
   Search Intent: [from research: informational / commercial / local / comparison / transactional]
   Audience: [inferred]
   Geography: [if relevant]
   Page Type: [from research: service page / listicle / comparison / pricing / local page / guide]
   Vertical: [airport parking / local service / SaaS / medical / legal / etc.]
   Information Gain Target: [what should this page add that the top 10 do not?]
   Reddit Test Target: [which subreddit? what would a knowledgeable commenter expect?]
   Word Count Target: [from research: recommended_min to recommended_max]
   H2 Target: [from research: median H2 count]
   PAA Questions to Answer: [from research]
   

   Confirm with user before writing unless they said "just write it."

3. Write: Front-load the fast-scan summary matrix in the first 200 words. Build 500-token chunks using the Snippet Answer rule. Integrate the "Not For You" block.

4. FAQ Section: Include a dedicated FAQ section answering at least 3 People Also Ask questions from research data. Each Q&A pair must be wrapped in FAQPage schema. This is NOT optional.

5. Hub & Spoke Links: If the page is a hub, list its spoke pages with links. If it's a spoke, link back to its hub. Include a "Related Pages" or "More Guides" section at the bottom with actual internal link targets.

6. Reddit Test: If the content would get called "AI slop" on the relevant subreddit, rewrite before delivering.

7. Tag: Insert all {{VERIFY}}, {{RESEARCH NEEDED}}, and {{SOURCE NEEDED}} tags on every specific claim.

8. Schema Markup: Generate complete JSON-LD schema block(s) at the end of the page. Required per page type (Section 6). Also embed key entities inline using RDFa or Microdata spans where appropriate. Do NOT skip this step.

9. Quality Checklist: Run the checklist (Section 14) and print the scorecard in the output (see Section 14 for format). If any item fails, revise before delivering.

10. Save: Output to ~/Documents/SEO-AGI/pages/ (new pages) or ~/Documents/SEO-AGI/rewrites/ (rewrites).

Rewrite Protocol

When rewriting an existing page:

1. Fetch URL (WebFetch) or read local file
2. Identify target keyword from title/H1 or ask user
3. Run research against the keyword
4. Run GSC data if available: for dir in "." "${CLAUDE_PLUGIN_ROOT:-}" "$HOME/.claude/skills/seo-agi" "$HOME/.agents/skills/seo-agi" "$HOME/seo-agi"; do [ -n "$dir" ] && [ -f "$dir/scripts/gsc_pull.py" ] && SKILL_ROOT="$dir" && break; done; python3 "${SKILL_ROOT}/scripts/gsc_pull.py" "<site_url>" --keyword="<keyword>"
5. Gap analysis: compare existing page vs research data. What's missing? What's thin? What fails the Reddit Test?
6. Rewrite following gap report
7. Output rewritten page + change summary (what changed and why)

Batch Mode

For batch requests ("write 5 location pages for [service]"), decompose into parallel sub-agents:

• Research agent: Run research per keyword variant
• GSC agent: Pull performance data if creds available
• Writer agent: Generate each page from its brief, following full execution protocol
• QA agent: Run quality checklist on each page

14. QUALITY CHECKLIST

Run before every delivery. If any answer is NO, revise before delivering.

MANDATORY: Print this scorecard at the end of every page output. Use the exact format below so the user can see what passed and what needs attention.

Pages scoring below 16/20 must be revised before delivery. Items marked NO must include a note on what needs to be fixed.

15. OUTPUT FORMAT

All pages output as Markdown with YAML frontmatter:

---