Semantic SEO Suite

A free, open-source alternative to Surfer, MarketMuse, Clearscope & Frase — built as Claude Code skills. Topical authority, content briefs, AI writing, schema, GEO, and backlinks. With a fabrication guard that refuses to invent numbers.

· Free forever · · No SaaS, no seat fees

Paid "AI SEO platforms" cost $50–$1,000+/month and quietly fabricate: made-up search volumes, invented "content scores," hallucinated product claims. This suite does the same jobs — entity-based topical mapping, content-gap analysis, keyword clustering, content briefs, on-page optimization, internal linking, schema markup, rank tracking, and answer-engine optimization — as skills you run inside Claude Code, for free, with the provenance of every number tagged and a guard that blocks any claim that isn't grounded.

Clone it, open the folder in Claude Code, talk to it in plain English.

---

Why this instead of a paid SEO tool?

	Semantic SEO Suite	Surfer / MarketMuse / Clearscope / Frase / SemanticOS
Price	Free, MIT, unlimited projects	$50–$1,000+/mo, per-seat, project caps
Where it runs	Your Claude Code — artifacts are files you own	Their cloud; you rent access
Topical map & authority	✅ pillar/cluster/supporting + prioritized calendar	✅ (their core feature)
Content briefs	✅ entity-aware, with a do-not-fabricate list	✅
Content optimization score	✅ three scores: fabrication, quality/depth, AEO	✅ one proprietary "content score"
Fabrication guard	✅ refuses invented stats & unverified brand claims	❌ no fact guard — AI output can hallucinate
AEO / GEO (get cited by ChatGPT, AI Overviews)	✅ dedicated skill + scorer	⚠️ partial / emerging
Off-page + distribution	✅ backlink prospecting + repurposing plans	❌ usually separate products
Keyword volume / SERP data	Bring-your-own (SerpApi/GSC) or run free at T0	Bundled proprietary database
Customizable	✅ edit the skills & rulebook yourself	❌ fixed black box

Where paid tools still win (honestly): they own huge proprietary keyword/SERP databases, polished dashboards, and one-click integrations. This suite trades the UI for transparency, ownership, and zero cost — and lets you plug in the same data via your own SerpApi/GSC keys.

What you can do

• Build topical authority — a full pillar → cluster → supporting topical map + prioritized content calendar from one central entity (Koray-style, entity-based).
• Model your niche as entities — Entity-Attribute-Value (EAV) architecture, not a flat keyword list, plus keyword clustering into topics.
• Generate content briefs — semantically ordered H2/H3 outlines with snippet targets, required entities, internal-link targets, and intent-conflict detection before you write.
• Write drafts that clear 3 gates — a fabrication guard, a depth/quality score, and an AEO citation score, every time.
• Audit any site — content-gap analysis, keyword cannibalization and entity-drift detection (via semantic distance / context vectors), orphan pages, and a bulk-URL crawl.
• Analyze competitors — model competitor entities and surface the topical + backlink gaps.
• Optimize internal linking & schema — hub-and-spoke link plans with varied, penalty-safe anchors + valid JSON-LD per page.
• Scale with programmatic SEO — entity-driven templates for pages at scale (e.g. per-vertical, per-integration).
• Track performance — Google Search Console rollups, striking-distance queries, decay, and uncovered-query candidates.
• Win AI answers (GEO) — harden content to be cited by ChatGPT, Perplexity, and Google AI Overviews.
• Grow off-page & distribute — backlink prospecting, competitor backlink-gap, and per-piece channel + repurposing plans.
• See it all on a visual gap heatmap — coverage, drafts, and gaps colour-coded at a glance.

Not included (yet — honestly)

No Google Business Profile / local-SEO module, no dedicated multilingual mode (it's language-agnostic at T0 but not specialized), and exports are Markdown / JSON / HTML rather than one-click Notion / Sheets / PDF. And you won't find a "+312% traffic" claim anywhere — we don't publish outcome stats we can't attribute. PRs welcome.

What you get (the actual deliverables)

Everything is a file you own, not a dashboard you rent:

• 🗺️ Topical map — topical-map.json + a readable tree (topical-map.md)
• 🔥 Visual coverage heatmap — a self-contained heatmap.html you open in any browser; color-codes every node by status (planned → briefed → drafted → published → needs-update)
• 📅 Prioritized content calendar — re-sorted on real SERP signals when you add a key
• 📝 Content briefs — per node, with entities, snippet target, and do-not-fabricate list
• ✍️ Guarded drafts — Markdown with front-matter, scored on fabrication / quality / AEO
• 🔗 Internal-linking plan — up/down/lateral links with varied, safe anchors
• 🏷️ JSON-LD schema — one validated file per page (Article/FAQ/HowTo/Product) + org graph
• 🔍 Audit report — gaps, cannibalization, drift, orphans, provenance-tagged
• 📈 Performance report — GSC rollups, striking-distance, decay
• 🤖 AEO report — per-page citation-readiness + a hardening queue
• 🎯 Link + distribution plans — prospects, anchors, channels, cadence

See a complete worked example in examples/driftroast/ — a fictional coffee brand with a real map, brief, guarded draft (scores: fabrication CLEAN · quality 100 · AEO 92), heatmap, schema, and link plan.

The 10 skills

Skill	What it does
seo-brand-foundation	Central entity, audience, competitors, EAV attributes, locked-facts ledger
topical-map-builder	Topical map + prioritized content calendar
content-brief-generator	Entity-aware brief: outline, snippet target, do-not-fabricate list
semantic-draft-writer	Writes the draft, runs 3 gates (fabrication → quality → AEO)
linking-and-schema	Internal-link plan + JSON-LD structured data
semantic-site-auditor	Content-gap, cannibalization, entity-drift, orphan audit
seo-performance-tracker	Google Search Console feedback loop
answer-engine-optimizer	GEO — get cited by ChatGPT / Perplexity / AI Overviews
link-opportunities	Backlink prospecting, competitor gap, safe anchor mix
content-distribution	Channels, repurposing atoms, cadence

The three quality gates (what replaces a fake "content score")

1. validate_draft.py — fabrication guard: no naked stats, no unbacked brand claims (hard gate).
2. draft_quality.py — depth, examples, specificity, anti-filler (the bar to beat).
3. aeo_score.py — citation-readiness for AI answer engines (a floor to clear).

Every number the suite emits is tagged measured, derived, or asserted. A bare number is treated as a bug.

Grounding tiers — free, or plug in data

Tier	Needs	Adds
T0	nothing	Full methodology on pure LLM + web search.
T1	free installs / GSC OAuth	Embeddings (cannibalization/drift), site crawl, Search Console.
T2	a SerpApi key	Live SERP signals → data-driven calendar prioritization.

Keys live only in a git-ignored .env (copy .env.example) — they never enter the repo.

Quickstart

git clone https://github.com/siddiqss/semantic-seo-suite.git
cd semantic-seo-suite

1. Install dependencies

Python 3.10+ required. T0 (the full methodology) needs almost nothing:

# T0 — required minimum (no accounts, no keys):
pip install jsonschema pyyaml

# T1 — optional (embeddings for cannibalization/drift, site crawl, Search Console):
pip install sentence-transformers trafilatura pytrends beautifulsoup4 \
            google-api-python-client google-auth-oauthlib

# T2 — optional (live SERP data): nothing to install — just add SERPAPI_KEY:
cp .env.example .env      # then paste your key into .env (git-ignored)

On an externally-managed Python (Homebrew/Debian) add --break-system-packages, or use a venv: python -m venv .venv && source .venv/bin/activate.

2. Use it in Claude Code

claude            # or open the folder in the Claude Code desktop app

Or install as a plugin (one command, no clone):

/plugin marketplace add siddiqss/semantic-seo-suite
/plugin install semantic-seo-suite

Either way, then just ask:

"Set up the SEO foundation for mydomain.com." · "Build the topical map." · "Write a brief for X, then draft it." · "Audit my site." · "How do we get cited by ChatGPT?"

The skills write everything into your brand workspace (brands/<your-slug>/).

3. Open the topical map & heatmap

The map ships as JSON + a readable Markdown tree; the heatmap is a self-contained HTML file you render, then open in any browser. Using the bundled demo:

# the topical map as a readable tree — open in any editor
open examples/driftroast/topical-map.md          # macOS   (Linux: xdg-open · Windows: start)

# render the visual coverage heatmap, then open it
python scripts/map_heatmap.py \
  --map examples/driftroast/topical-map.json \
  --out examples/driftroast/audits/heatmap.html
open examples/driftroast/audits/heatmap.html     # macOS   (Linux: xdg-open · Windows: start)

For your own brand, swap examples/driftroast → brands/<your-slug>.

Try it now (offline scorers + self-tests)

# fabrication guard on the demo draft → CLEAN
python scripts/validate_draft.py \
  --draft examples/driftroast/drafts/how-to-make-pour-over-coffee.md \
  --locked examples/driftroast/locked-facts.json --brand-terms "DriftRoast"

# quality + AEO scorers
python scripts/draft_quality.py --draft examples/driftroast/drafts/how-to-make-pour-over-coffee.md --floor 700
python scripts/aeo_score.py     --draft examples/driftroast/drafts/how-to-make-pour-over-coffee.md

# every script's self-test
for s in aeo_score draft_quality link_prospects distribution_plan reprioritize_map serpapi_client provenance; do python scripts/$s.py --selftest; done

Full instructions per tier: RUNBOOK.md. Verify the suite yourself: TESTING.md.

Use it in the Claude apps (Desktop / web)

Not a developer? Run the skills in the Claude Desktop app — no clone, no CLI.

Claude Desktop (easiest): in a chat, click the + icon → Add plugin → paste the repo URL:

https://github.com/siddiqss/semantic-seo-suite

Claude reads the plugin, shows the 10 skills, and installs them. Then just ask — "Build a topical map for mydomain.com," "write a brief and draft it," "how do we get cited by ChatGPT?"

Alternative — upload skill packages (e.g. claude.ai web, or org setups without plugin install): download the individual .skill files from the latest release and add each under Settings → Capabilities → Skills. (Custom Skills is a Pro/Max/Team/Enterprise feature.)

Either way, the apps run the skills' reasoning and lighter scripts; the heavy steps (site crawl, embeddings, Search Console) need Claude Code, which executes them directly against a real filesystem. Same skills.

Repo layout

skills/       10 Claude Code skills (SKILL.md each)
framework/    14 methodology docs the skills read
scripts/      25 deterministic helpers (guards, scorers, SERP/GSC clients) — most ship --selftest
templates/    JSON schemas + config template
examples/     driftroast — a complete, fictional worked demo brand
brands/        your workspaces (git-ignored — client data never committed)

About

Built by Talha Siddique — Senior Full Stack Engineer and I build grounded, no-nonsense production AI systems. This suite is free and MIT-licensed; use it for your own brand, client work, or ventures. Issues and PRs welcome; star it if it saved you a subscription.

License

MIT.