Pipeline Position
Step 3 of 4 — consumes morphiq-rank output.
- Input: Prioritized Roadmap (JSON) OR user prompt OR existing content.
- Output: Build Output (JSON + artifacts) → consumed by morphiq-track.
- Data contract: See
PIPELINE.md§3 for the Build Output schema.
Purpose
Morphiq Build fixes issues from morphiq-rank. It creates new content, optimizes existing content, generates schema markup, builds policy files, and produces artifacts to improve AI visibility. The core engine is a 6-step content lab pipeline.
Entry Points
Path A — From Prioritized Roadmap: Process issues by tier and priority. Route each to the appropriate fix workflow.
Path B — From User Prompt: Accept topic, optional source URLs (up to 5), optional ICP/brand context. Route to content lab pipeline.
Path C — From Existing Content: Accept content URL or raw text. Route to quality rewrite workflow.
Content Lab Pipeline (6 Steps)
Step 1: Ingest Sources
Validate URLs, filter blocked domains, deduplicate, cap at 10. Accept raw text or PDF alternatives. Halt if zero valid sources.
Step 2: Extract Content
Crawl each URL → clean markdown. Extract title, content, outbound links, publish date. Halt if zero successful extractions.
Step 3: Analyze Gaps
Analyze against query space. Identify 5 gap types:
| Gap Type | What Is Missing |
|---|---|
| Content | Unanswered questions, missing perspectives |
| Data | Missing statistics, quantitative evidence |
| Format | Wrong format for LLM retrieval |
| Depth | Surface-level, no expert insight |
| Fanout coverage | Sub-queries AI would chain but site cannot answer |
Detect comparative intent. Evaluate fanout coverage using content type → sub-query rules. Generate up to 5 search queries.
For gap taxonomy and severity, read references/gap-taxonomy.md.
Step 4: Research to Fill Gaps
Run up to 5 live web searches. Collect authoritative sources, statistics (number + source + URL), expert quotes (speaker + credential), industry insights. If comparative intent, dedicate 1 search to brand data.
For citation rules, read references/enrichment-sources.md.
Step 5: Generate / Rewrite
Produce final content applying Morphiq standard:
- E-E-A-T signals, name-drop citations, expert quotes
- Heading hierarchy, 50–75 word paragraphs, direct-answer blocks
- Brand positioning (comparative or authority mode)
- 1,200–1,600 words, 5–7 H2 sections, FAQ with 3–5 Q&As
- Minimum 3 statistics, 1 expert quote, sources section
- No fabricated case studies
For full pipeline spec, read references/content-lab-pipeline.md.
Step 6: Validate Fanout Coverage (Fanout Issues Only)
For fanout-* issues with fanout_context: validate generated content addresses all triggering sub-queries and meets the competitive quality floor. If coverage < 80% or quality floor not met, revise once. Skip for non-fanout content.
Run scripts/validate-coverage.py with generated content + triggering sub-queries + quality floor.
Post-Pipeline Processing
| Process | What It Does | Reference |
|---|---|---|
| Schema Injection | Classify content type, generate JSON-LD. New content: embed schema in content artifact. Existing content: separate schema artifact with implementation tracking. | references/schema-templates.md |
| Metadata Optimization | Meta description, slug, OG tags | references/metadata-patterns.md |
| llms.txt Generation | Full autonomous pipeline: scrape → LLM → validate → repair → template fallback | references/llms-txt-spec.md |
| Content Restructuring | Fix headings, split paragraphs | — |
| Internal Linking | Link related pages for site: coverage | references/content-lab-pipeline.md |
| Enrichment | Additional search for missing stats/citations | references/enrichment-sources.md |
| FAQ Generation | Generate FAQ from gap analysis | references/faq-guidelines.md |
Issue Type → Fix Routing
| Issue Category | Fix Approach |
|---|---|
agentic-* schema | Schema Injection — generate JSON-LD |
agentic-* metadata | Metadata Optimization — generate tags |
content-* quality | Quality Rewrite — Step 5 pipeline |
chunking-buried-answer | Quality Rewrite — Step 5 pipeline (Claude-driven rewrite to answer-first structure) |
chunking-* structure (other) | Content Restructuring |
policy-* files | Policy file generation |
fanout-* coverage | Full 6-step pipeline for new content. When fanout_context is present, pass triggering_sub_queries to Step 3 and competitor_sources to Step 4. Run Step 6 (coverage validation) before post-pipeline processing. |
visibility-* | Enrich existing content via pipeline |
Build Output
Artifacts with type: "content", "schema", "metadata", "policy_file". Each includes placement instructions.
Reference Files
| File | Purpose |
|---|---|
references/content-lab-pipeline.md | Full 6-step pipeline with I/O formats |
references/gap-taxonomy.md | Gap types, severity, search query rules |
references/enrichment-sources.md | Citation format, source preferences |
references/schema-templates.md | JSON-LD templates, skip conditions |
references/metadata-patterns.md | SEO metadata rules |
references/llms-txt-spec.md | llms.txt spec and generation |
references/faq-guidelines.md | FAQ generation rules |