# Topical Authority Playbook
**Portable spec for building SEO + GEO topical authority for any new SaaS app.**
Built from the Kompozy / BILT KONTENT ENGINE buildout (2026-04-23 → 2026-05-20). Reusable across the 10-app portfolio: cold email engine, cold SMS, cold DMs, cold WhatsApp, affiliate platform, giveaway app, lead scraper, AI cold caller, and any future SaaS.
Hand this spec to a Claude Code agent on a new app project. Tell the agent: **"Execute this playbook against [APP_NAME] in [DOMAIN], ICP is [ICP], primary keyword cluster is [KEYWORD]."** The agent will know exactly what to build.
---
## 0. Use this spec
**Inputs the agent needs from the user before starting:**
1. **APP_NAME** — e.g. "Kompozy"
2. **DOMAIN** — e.g. "kompozy.io" (must be registered + Vercel-attached before the agent runs)
3. **ONE_LINER** — e.g. "Autonomous content engine — ingest one source, fan out to 5 output buckets, publish to 9 platforms."
4. **ICP** — specific buyer (e.g. "real estate agents", "B2B SaaS founders running outbound", "creator-operators with podcast + newsletter")
5. **PRIMARY_KEYWORD_CLUSTER** — 1-3 head keywords (e.g. "AI content tools", "cold email automation", "lead scraping")
6. **TOP_5_COMPETITORS** — exact brand names + URLs (e.g. OpusClip, Submagic, HeyGen)
7. **PRICING_TIERS** — name + monthly price + credits/seats per tier
8. **FOUNDER_BIO** — 2-3 sentences (used in `Person` schema + About page)
**Outputs the agent will produce:**
- 250-350 prerendered SEO pages on the marketing surface
- 15-25 free interactive tools at `/tools`
- Full schema.org graph (Organization, Article, FAQPage, HowTo, SoftwareApplication, Person, BreadcrumbList)
- Sitemap + robots.txt + llms.txt
- Email auth recipe (SPF + DKIM + DMARC)
- A `docs/outreach/seo-geo-todos.csv` operator checklist
- A weekly LLM citation tracker (optional — Phase 3)
Total agent time: **~40-60 hours of focused build work** depending on app complexity, spread across 4-6 weeks calendar time. Founder time: **~30-50 hours** (mostly outreach, reviews, content review).
---
## 1. The 5-layer topical authority model
Every layer reinforces every other layer. Don't skip layers.
```
┌────────────────────────────────────────────────────────┐
│ L5: Research │ Original data, "Index" pieces (citation magnets)
│ Examples: AI Tells Index, deliverability benchmarks │
├────────────────────────────────────────────────────────┤
│ L4: Tools (free, indexable) │ Pure-math calculators + LLM-powered tools
│ Examples: Stack Cost Comparison, AI Tells Detector │
├────────────────────────────────────────────────────────┤
│ L3: Spokes │ /guides, /roundups, /migrate, /alternatives,
│ │ /workflows, /outputs, /integrations, /for, /blog
├────────────────────────────────────────────────────────┤
│ L2: Cluster Hubs │ 10-12 pillar pages, one per topic family
│ Examples: /repurpose, /autonomous, /brand-voice │
├────────────────────────────────────────────────────────┤
│ L1: Homepage + Brand Surface │ Single page, full Organization schema
└────────────────────────────────────────────────────────┘
```
**Why this stack:** Google + LLMs both reward topical depth. A single 5,000-word post on "AI content tools" loses to 100 pages where every page covers one specific facet of the topic and they all link to each other.
**Cluster count discipline:** 10-12 cluster hubs is the sweet spot. Fewer = thin authority. More = diluted attention; you can't service 30 clusters well as a small team.
---
## 2. File / folder architecture
Mirror this exactly in the new app's Next.js App Router project. Paths shown are `src/app/...`.
```
src/app/
├── page.tsx # Homepage — Organization schema, full ICP messaging
├── about/page.tsx # Organization + Person schema (founder)
├── pricing/page.tsx # Pricing tiers + Offer schema
├── pricing/calculator/page.tsx # Interactive pricing calculator (tool)
├── signup/page.tsx # Conversion endpoint, NoIndex meta
├── compare/page.tsx # Master competitor comparison grid
│
│ # ─────── L2: Cluster hubs (one folder per cluster) ───────
├── [cluster-1-slug]/page.tsx # e.g. /repurpose, /autonomous, /brand-voice
├── [cluster-1-slug]/[slug]/page.tsx # Cluster spokes (10-15 per cluster)
├── [cluster-2-slug]/...
│ # ... 10-12 total cluster hubs ...
│
│ # ─────── L3: Spokes (cross-cutting page types) ───────
├── guides/page.tsx # Index of long-form guides
├── guides/[slug]/page.tsx # Article schema, TOC, direct-answer, FAQ
├── roundups/page.tsx # "Best [X]" listicle index
├── roundups/[slug]/page.tsx # Article + ItemList + FAQPage schema
├── migrate/page.tsx # Migration guides index
├── migrate/from-[competitor]/page.tsx # HowTo + FAQPage schema
├── alternatives/page.tsx # Alternatives hub
├── workflows/[slug]/page.tsx # Source → destination workflow pages
├── outputs/[slug]/page.tsx # One per output type your product produces
├── integrations/[slug]/page.tsx # One per 3rd-party integration
├── for/[slug]/page.tsx # ICP / industry-specific landing pages
├── blog/page.tsx # Blog hub
├── blog/[slug]/page.tsx # Article schema (NO opengraph-image.tsx — see §10)
├── research/page.tsx # Original-data studies index
├── research/[slug]/page.tsx # The citation-magnet pieces
├── reviews/page.tsx # Customer reviews aggregator
├── case-studies/page.tsx # Named customer stories
├── glossary/page.tsx # Term index
├── glossary/[slug]/page.tsx # DefinedTerm schema
├── faq/page.tsx # FAQPage aggregator
├── changelog/page.tsx # Product update log
│
│ # ─────── L4: Tools (free, indexable) ───────
├── tools/page.tsx # Tool gallery index
├── tools/[tool-slug]/page.tsx # SoftwareApplication + HowTo schema
│
│ # ─────── L5: Research ───────
├── research/[slug]/page.tsx # Same path as L3 — same data study layer
│
│ # ─────── Discoverability files ───────
├── sitemap.ts # Generates from src/lib/* manifests
├── robots.ts # Allow / + disallow /dashboard, /admin, /api
├── llms.txt/route.ts # LLM-readable site map
├── opengraph-image.tsx # ROOT ONLY — never per-slug (see §10 lab note)
├── icon.svg # Brand favicon
├── apple-icon.png # 180x180 PNG (NOT SVG — iOS rejects)
└── not-found.tsx # 404 with noindex meta
```
**Library / data files** (where content lives, separate from page components):
```
src/lib/
├── clusters.ts # Cluster hub definitions (slug, hubTitle, faq, spokes[])
├── spokes.ts # Spoke posts under each cluster
├── use-cases.ts # /for/[slug] ICP pages
├── seo-collections.ts # TEMPLATES, OUTPUTS, WORKFLOWS, INTEGRATIONS arrays
├── guides.ts # Long-form guide entries
├── roundups.ts # "Best [X]" listicle entries
├── migrations.ts # Per-competitor migration guides
├── vs-comparisons.ts # Comparison page data
├── glossary.ts # Glossary terms
├── blog-posts.ts # Blog post entries (BlogPost type)
├── tools-manifest.ts # Tools index (status: 'live' | 'soon')
├── glossary-autolink.ts # Auto-link first occurrence of glossary terms
└── research/[study-slug].ts # One file per data study
```
**Shared components:**
```
src/components/
├── MarketingPage.tsx # Outer chrome (header, footer, nav)
├── ClusterHubPage.tsx # L2 hub layout (renders hub + cluster spokes)
├── ClusterSpokePage.tsx # L3 spoke layout
├── ToolShell.tsx # Tool page wrapper (JSON-LD + breadcrumbs + CTA)
├── Breadcrumb.tsx # Single breadcrumb component used everywhere
└── EmbeddedVideo.tsx # YouTube embed + VideoObject schema (optional)
```
---
## 3. Schema discipline by page type
**Every page MUST emit at least one schema.org type.** Pages without schema get ignored by LLM citation systems.
| Page type | Required schemas | Required fields |
|---|---|---|
| Homepage | `Organization` + `WebSite` + `BreadcrumbList` | `name`, `url`, `logo`, `sameAs[]` (LinkedIn, X, YouTube, GitHub, Crunchbase), `founder`, `description` |
| About | `Organization` + `Person` (founder) + `BreadcrumbList` | `Person.name`, `Person.url`, `Person.image`, `Person.sameAs[]`, `Person.affiliation` |
| Cluster hub | `WebPage` + `BreadcrumbList` + `FAQPage` (5-7 FAQs) | `name`, `description`, FAQ Q&A pairs |
| Cluster spoke | `Article` + `FAQPage` + `BreadcrumbList` | `headline`, `author` (Person), `datePublished`, `dateModified`, `publisher` (Organization) |
| Guide / blog post | `Article` + `FAQPage` + `BreadcrumbList` | Same as spoke. Add direct-answer 40-80 word paragraph for LLM citation. |
| Roundup | `Article` + `ItemList` + `FAQPage` + `BreadcrumbList` | ItemList items name each competitor with `Product` schema if appropriate |
| Migration guide | `HowTo` + `FAQPage` + `BreadcrumbList` | Numbered steps with `HowToStep.name` + `HowToStep.text` |
| Comparison page | `Article` + `BreadcrumbList` | Mention competitor as `Product` if defensible |
| Tool page | `SoftwareApplication` + `HowTo` + `FAQPage` + `BreadcrumbList` | `name`, `applicationCategory`, `operatingSystem`, `offers` (Price: $0) |
| Research / data study | `Article` + `Dataset` + `BreadcrumbList` | `Dataset.name`, `Dataset.distribution[]` (link to Hugging Face if published) |
| Reviews | `CollectionPage` + `BreadcrumbList` | **Only emit `AggregateRating` + `Review` schema when you have 3+ NAMED, verifiable reviewers.** Google penalizes anonymous review markup. |
| FAQ | `FAQPage` + `BreadcrumbList` | One FAQPage per page; merge into Article schemas where appropriate |
| Glossary term | `DefinedTerm` + `BreadcrumbList` | `name`, `description`, `inDefinedTermSet` |
**JSON-LD emission pattern (consistent across the codebase):**
```tsx
const jsonLd = {
'@context': 'https://schema.org',
'@graph': [
{ '@type': 'Article', '@id': `${url}#article`, ... },
{ '@type': 'FAQPage', mainEntity: [...] },
{ '@type': 'BreadcrumbList', itemListElement: [...] },
],
};
return (
<>
{/* page content */}
);
```
Always inline (`dangerouslySetInnerHTML`), never use `