pSEO Notebook Research — Building and Scaling Programmatic SEO Systems

Source: NotebookLM notebook c46a44df-8628-4fcf-9a0a-5a4918a736b5 (50 sources) Queried: March 2026 | Status: PARTIAL — 3 of ~8 planned queries completed Next session: re-query remaining topics AND save each as NotebookLM note

---

Q1: URL Structure — /blog vs /learn vs flat

Decision: Use /learn/ for GrowReach (not /blog/)

• Flat structure DISCOURAGED — dilutes authority, Google can't map topic relationships
• Hub-and-spoke nested structure wins: central hub → multiple spokes → cross-links back
• /learn/ or /resources/ signals educational hub — better topical authority positioning than /blog/
• Successful SaaS directories: /use-cases/, /integrations/, /templates/, /learn/

Recommended GrowReach URL silos:

• /learn/[slug] — informational SEO articles (CFBR, engagement guides, etc.)
• /free-tools/[tool-name] — free tool pages
• /use-cases/[slug] — use case pages
• /compare/[growreach-vs-competitor] — comparison pages
• /features/[slug] — product feature pages

Technical requirements:

• Every page reachable within 3 clicks of homepage (crawl efficiency)
• Breadcrumb schema reinforces hierarchy
• Segmented sitemaps by category (not one giant sitemap)
• Canonical tags prevent duplicate URL penalties
• llms.txt at root guides AI crawlers (ChatGPT, Perplexity, Claude)

---

Q2: Complete pSEO Workflow — 7 Phases

Phase 1: Keyword Research

• Find scalable PATTERNS: head term + modifiers
• GrowReach pattern: "linkedin comment" + [automation / generator / examples / for sales / for founders]
• Tools: SemRush Keyword Magic Tool (user-operated) + NeuronWriter for SERP depth
• Output: keyword matrix — each row = one page

Phase 2: Unique Data Collection

• Pure AI content without unique data = Google penalty risk
• GrowReach unique data: LinkedIn algorithm stats, engagement benchmarks, comment examples by industry/role
• This is our competitive moat — competitors can't replicate our data layer

Phase 3: Keyword Matrix (Google Sheets)

• Columns: keyword, slug, H1, meta title, meta description, intro variable, template type
• Each row = one published page
• Airtable for relational data if needed (update one field → updates all pages)

Phase 4: Content Generation

• NeuronWriter API → NLP terms per keyword → passed as mandatory inclusions to Claude
• Claude writes unique content per page using NLP terms + matrix data
• Dynamic prompts (not copy-paste) = genuinely unique content

Phase 5: Publishing to WordPress

• REST API: POST /wp-json/wp/v2/posts
• Fields: title, content (HTML), slug, status=draft, categories, meta (via Rank Math API Manager)
• 1-2 second delays between posts (server bottleneck protection)
• Always draft first — human review before publish

Phase 6: SEO + GEO Optimization (every page)

• JSON-LD schema: FAQPage + Article + Organization
• llms.txt at root
• Rank Math meta fields via Rank Math API Manager plugin (extends REST API)
• All text in initial HTML — no JS-rendered content (AI crawlers can't read it)

Phase 7: Monitoring & Scaling

• Indexation ratio target: >60%
• Engagement within 30% of site average
• Scale gradually: 20-30% more pages/month
• Tools: Google Search Console + GA4

---

Q3: Claude Code + WordPress REST API Automation

Core insight: Claude Code is not just a writer — it's a terminal-based orchestrator that can manage the entire pSEO pipeline end-to-end via MCP and REST API.

Authentication

• WordPress Application Password: Users > Profile > Application Passwords
• Already configured in CREDENTIALS.md: username=rahul, password=HQss dyby ZbFv Qz6p hQ7h jEHi
• REST API base: https://growreach.app/wp-json/wp/v2/

WordPress fields set per published page

Standard fields (REST API native):

• title: SEO-friendly H1
• content: AI-generated HTML body
• status: draft (always — human review gate)
• slug: clean keyword-rich URL
• categories: numerical IDs for topic silos

SEO plugin fields (require Rank Math API Manager plugin):

• rank_math_title → SEO title tag
• rank_math_description → meta description
• rank_math_focus_keyword → primary keyword
• rank_math_canonical_url → canonical

The automation script pattern (scripts/wp-publish.js — TO BUILD)

1. Read keyword matrix row (from Google Sheets or local CSV)
2. Get NeuronWriter NLP terms for the keyword
3. Claude generates HTML article using NLP terms
4. POST to /wp-json/wp/v2/posts with all fields
5. POST to /wp-json/rank-math-api/v1/update-meta with SEO fields
6. Wait 1-2 seconds
7. Next row

Key warning from sources

• Server bottleneck is the #1 failure point — Claude generates faster than WordPress can accept
• Implement 1-2 second delays between API calls
• Test with 1 page first before running bulk
• One case study: Claude Code successfully pushed 460 posts in bulk using this pattern

---

REMAINING QUERIES TO RUN NEXT SESSION (save each as note)

1. Topical authority + hub-and-spoke content model for SaaS — how many pages needed?
2. GEO (Generative Engine Optimization) — how to get cited by ChatGPT/Perplexity/AI Overviews
3. WordPress template structure for pSEO — technical requirements, schema, Core Web Vitals
4. Internal linking strategy at scale — how to automate it
5. NeuronWriter API deep integration with Claude Code workflow

---

NEXT SESSION INSTRUCTIONS

1. Open notebook c46a44df-8628-4fcf-9a0a-5a4918a736b5
2. Run each remaining query above ONE AT A TIME
3. IMMEDIATELY save each response as a NotebookLM note (mcp__notebooklm-mcp__note action=create)
4. After all queries done, synthesize into GrowReach pSEO Master Plan
5. Present plan to user for review before any execution