lighthouse-index.com

The Challenge

Cataloguing thousands of cultural landmarks with unique, factually accurate, SEO-optimized content is not something you do by hand. But throwing raw LLM output at a website produces generic slop. The real challenge was engineering a pipeline where AI generates quality content at scale while a human-designed system controls structure, validates facts, handles geo-data, and optimizes for search engines.

My Contribution

Engineered a multi-stage content pipeline: prompt templates generate structured content per landmark, a validation layer checks facts against geographic databases, and a transformation step produces SEO-optimized pages with structured data (JSON-LD), internal linking, and semantic HTML.

Built the platform on Next.js with static site generation, serving four interconnected sites (lighthouse-index.com, castle-index.com, mill-index.com, church-index.com) from a single monorepo. I use AI as a content accelerator, but every architectural decision, from the data model and URL structure to the rendering pipeline and deployment strategy, is hand-crafted engineering.

Key Results

• Four index sites with thousands of unique, validated landmark pages live and indexed
• Strong organic search visibility, with pages ranking on the first page for long-tail queries
• Single monorepo architecture serving four distinct sites with shared infrastructure
• AI content pipeline with automated fact-checking against geographic databases

Technologies Used