Niche Hunter — Super-Niche Discovery & Article Mapping
Niche Hunter
TL;DR: Niche Hunter finds a “super-niche” worth dominating, proves it’s actually winnable before you commit writing budget, and hands back a 50-200 article map that exhausts it. Four phases: hypothesis generation → validation → article mapping → (optional) drafting. Uses Google autocomplete + browser-based SERP reading — no paid APIs.
Name note (October 2026): internally the skill is now called product-article-map, to tell it apart from a separate demand-measurement skill. The method on this page is unchanged.
What It Does
Most SEO advice says “write about your industry.” That loses every time against incumbents with ten years of backlinks. Niche Hunter implements the opposite play:
Pick a territory narrow enough to own in 3-6 months, then exhaust it with interrelated articles.
The skill operationalizes what SEO experts talk about but rarely execute: glossary/topical-authority through systematic niche selection.
The Core Concept: Super-Niche
A glossary/super-niche isn’t a keyword. It’s an intersection:
Audience × Problem × ContextBad examples:
- “Productivity” — war zone, you’ll lose
- “Keto” — millions of competitors
- “Content marketing for SaaS” — a decade of entrenched players
Good examples:
- “Productivity for ADHD remote workers in creative fields”
- “Keto for women 50+ managing thyroid issues”
- “Real estate investing in US Midwest tertiary markets under $100k”
The skill’s job: find that intersection, validate it’s winnable, map the 50-200 articles that exhaust it.
The Four Phases
Phase 1: Niche Hypothesis Generation (~5 min)
Input: Product URL or description
Process:
- Build product model (what it does, who buys, unique strengths, weaknesses, adjacent problems)
- Generate 5-10 candidate super-niches as
audience × problem × contextintersections - Each candidate gets: name, definition, audience, 3-5 seed queries, hypothesis for why it’s under-served
Output: niche_candidates.json
Phase 2: Five-Axis Validation (~15-25 min per candidate)
This is where most content strategies fail. Niche Hunter validates each candidate against five axes:
| Axis | Question | Method |
|---|---|---|
| Size | Enough query surface for 50-200 articles? | Google autocomplete expansion |
| Competition | Fragmented (takeable) or dominated? | SERP reading via browser |
| Commercial Density | Does audience actually buy the product? | LLM judgment on query intent |
| Expertise Fit | Can team credibly own this niche? | E-E-A-T signal assessment |
| AEO Gap | Do LLMs give shallow/hedged answers? | Probe 2-3 LLMs with 5-8 questions |
Verdicts: GO / MAYBE / SKIP with rationale per candidate
Output: niche_validation.json + recommendation of 1-2 candidates to commit to (or “skip all” if none pass)
Phase 3: Article Map Generation (~20-40 min)
Only runs after committing to a specific niche.
Process:
- Exhaustive expansion via autocomplete + PAA trees + Related searches
- Target: 300-1000 candidate queries, deduped
- Cluster into 80-200 article-sized units
- Assign roles: Pillar (5-10), Cluster (40-80), FAQ (20-40), Glossary (10-20)
- Prioritize into three build phases
Output: niche-<slug>-article-map-<date>.md
Phase 4: Article Drafting (Optional)
Only runs when explicitly requested. Turns map entries into publishable markdown with:
- Full frontmatter aligned with wiki schema
- Schema.org JSON-LD markup
- Internal wikilinks
- Role-specific format (glossary 300-600 words, pillar 2,500-5,000 words)
The Five-Axis Validation Rubric
Axis 1: Size
| Unique Candidates | Verdict |
|---|---|
| <30 | Fail — too narrow, starvation |
| 30-80 | Soft pass — minimum viable |
| 80-200 | Pass — sweet spot |
| 200-500 | Soft pass — borderline broad |
| >500 | Fail — too broad, not a niche |
Axis 2: Competition Shape
| SERP Pattern | What It Means |
|---|---|
| Fragmented | Mix of forums, Reddit, small blogs. No single domain in top 10 twice. Takeable. |
| Mixed | One authority + weaker results. Takeable but slower. |
| Dominated | 3+ spots owned by one major player (Healthline, NerdWallet, etc.). Skip. |
Axis 3: Commercial Density
High signals: comparison queries, pain-explicit queries, buyer-stage language Low signals: “what is X” definitions, “free X” requests, student/hobbyist audience
Axis 4: Expertise Fit
Can the team produce content with real authority? Signals:
- Named humans with credentials
- First-party data (“we analyzed 12,000 customers”)
- Existing case studies in the niche
- Podcast/YouTube presence
Axis 5: AEO Gap
Probe LLMs with representative questions. Wide gap = models hedge, refuse, or cite scattered sources. Narrow gap = models answer confidently with consistent citations from one authority.
Article Map Structure
Four-Tier Article Roles
| Role | Count | Length | Purpose |
|---|---|---|---|
| Pillar | 5-10 | 2,500-5,000 words | Canonical references for sub-topics |
| Cluster | 40-80 | 1,200-2,500 words | Mid-depth articles under pillars |
| FAQ | 20-40 | 400-900 words | Question-shaped, AEO-optimized |
| Glossary | 10-20 | 300-600 words | LLM-citation magnets |
Three Build Phases
| Phase | Timeline | Focus |
|---|---|---|
| Phase A | Weeks 1-4 | Quick wins (weakest SERPs, FAQ + glossary) |
| Phase B | Weeks 5-12 | Authority core (pillars + their clusters) |
| Phase C | Months 4+ | Completion (exhaust the niche) |
Real Example: Primores.org Test Run
Three candidates tested on Primores’ own positioning:
| Candidate | Verdict | Why |
|---|---|---|
| AI creative reverse-engineering | GO | Maps to existing IP (ad-alchemy), narrow framing is wide open |
| AI TikTok/IG distribution | SKIP | Product-intent SERPs, settled AEO, wrong audience |
| AI for marketing agencies | MAYBE | Winnable but audience doesn’t match current buyers |
Key catch: The skill identified that “AI meta ads” as a broad framing was being colonized by AdStellar.ai (5-6 of top 10 results). But the narrow “reverse-engineering as workflow” framing was wide open. That distinction would have wasted months if missed.
When to Use
- “Find me content niches for my product/service”
- “Plan articles for a domain”
- “Build topical authority in a space”
- “Research low-competition keyword clusters”
- “Do content-strategy groundwork”
Technical Requirements
- Google autocomplete (public endpoint, no API key)
- Browser-based SERP reading (Chrome MCP or similar)
- LLM access for AEO gap probing
- No paid keyword tools required
- Optional: measured search volume and difficulty can come from a separate demand-measurement run (DataForSEO), still experimental, instead of autocomplete estimates
Key Takeaways
- A super-niche is
Audience × Problem × Context— narrow enough to own, broad enough to sustain - Five-axis validation catches three failure modes: too broad, too narrow, wrong audience
- Article map is a graph, not a list — internal linking creates topical authority
- Build in phases: quick wins first, then authority core, then completion
- A “skip all” recommendation is more valuable than picking the least-bad niche
Related
- glossary/super-niche — The core concept
- glossary/topical-authority — The SEO strategy this implements
- seo/articles-engine — The operational loop this skill’s article map feeds (stage 1 of six)
- seo/agentic-search-optimization — How this fits into modern SEO
- tools/reddit-thread-analyzer — Complementary tool for sourcing content
- glossary/geo-aeo — Why AEO gap matters
Sources
- Primores internal skill (niche-hunter/SKILL.md)
- Test run on primores.org (2026-04-24)