Turn Your LLM Into a Pedantic SEO Auditor

What it does The repo is a two-part kit: a curated knowledge base of Google’s official SEO guidance split into nine topics, plus a self-executing prompt workflow that turns AI coding assistants into autonomous auditors. You point it at your codebase; it maps every public route, checks each page against Google’s rules, and generates a prioritized fix list with citations back to the source docs. The goal is to replace skimming a lengthy manual with an AI that actually runs the audit-and-fix loop on your specific routes.

The interesting bit Instead of yet another static checklist, this is an executable methodology: a five-phase prompt sequence (START.md and prompts/) that forces the AI to bootstrap your stack, discover routes, audit, prioritize infrastructure issues first, and verify fixes through builds. It was forged from a real audit of a 40-locale, 574K-concurrent-user production site, so the workflow is battle-tested rather than theoretical.

Key highlights

• Nine focused knowledge-base topics covering metadata, hreflang, structured data, Core Web Vitals, and 2024–2026 Google updates.
• Every audit rule cites its source in Google Search Central or web.dev, so you can verify the logic instead of trusting a blog post.
• The workflow classifies routes as public-index, public-noindex, or private before auditing, so it doesn’t waste time on admin panels.
• Output templates (ROUTES-INVENTORY.md, SEO-AUDIT-PROGRESS.md) give you a paper trail of what the AI checked and fixed.
• Explicitly distinguishes actual ranking factors from accessibility-only niceties, and logs deliberate skips for transparency.

Caveats

• The README admits that feeding the full knowledge base to a chat assistant only works “if it fits,” so large sites may bump against context-window limits.
• This is structured documentation and prompt engineering, not a standalone CLI tool; if your LLM struggles with long multi-step reasoning, the audit quality will vary.

Verdict Worth bookmarking if you maintain a content-heavy or multi-locale web app and want your AI assistant to handle tedious SEO hygiene. Skip it if you are looking for a traditional CLI scanner or expect magic without supervising the LLM’s output.