Skip to content

work / seo-graph · Live · real client work

SEO Graph

My own SEO lab: keyword graphs to client-ready reports.

Week of Claude Maxing

Play with it

A pure-frontend replica with mock data: nothing is saved, nothing leaves the page. The real app is richer; this is the feel of it.

loading the demo…

The problem

Agency-grade SEO tooling is expensive, and almost none of it treats AI search (AI Overviews, ChatGPT, Perplexity) as a first-class citizen.

What it does

A research engine that crawls autocomplete and Google Search Console, builds a keyword co-occurrence graph with cluster detection, surfaces content gaps, runs 32-check page audits with AI-search scoring, and renders branded, client-ready site reports. An AI-visibility panel measures share of voice inside grounded LLM answers.

Under the hood

  • Seven-category health score with an impact-by-effort matrix, in the report format my agency ships to clients
  • An InfraNodus-style physics graph for exploring keyword clusters and gaps
  • Produces the actual deliverables PTK Digital Labs sells: this tool pays for itself

The thinking

Target user

PTK Digital Labs first (me as the operator), and by design any freelance SEO or small agency that cannot justify enterprise tool subscriptions.

Pain points

  • Agency-grade keyword and audit tooling costs hundreds a month
  • Classic SEO tools treat AI search (AI Overviews, ChatGPT, Perplexity) as an afterthought
  • Deliverables still need hours of manual assembly after the research

Gap in the market

Nothing combined graph-based keyword research, a weighted audit engine and AI-search visibility measurement in one tool that outputs a branded, client-ready report.

AI + tooling

  • Gemini (enrichment + grounded AI-visibility checks)
  • Claude Code (build)
  • Google Search Console API
  • Louvain clustering
  • Cytoscape.js
  • FastAPI + Supabase

Frameworks

  • Network theory: keyword co-occurrence graphs surface structure lists cannot
  • Topical authority + content-gap theory for what to write next
  • Search-intent classification (informational, commercial, transactional)
  • E-E-A-T content lens + AIEO scorecard (own)
  • 7-category weighted health score with impact by effort matrix

Success metrics

  • Produces the actual PTK client deliverables
  • Full-site report from crawl to PDF in minutes
  • TBD: paid report count once productised
what I learned
Serverless has no disk. Porting SQLite to Postgres behind one dialect layer beat a rewrite.
next projectGymMate