Glow Index
An AI-powered beauty intelligence engine using Next.js, Llama 3, and Reddit data to transform unvarnished community discussions into structured product sentiment.
Summary
I built Glow Index (GlowIndex.com) to answer a simple, recurring question: What does Reddit actually think about this beauty product?
Not what a sponsored influencer thinks. Not what the product description says. Not a five-star review from someone who received the item for free. What are real people saying when they talk to each other?
Powered by Llama 3, Next.js, and Reddit data, Glow Index turns thousands of messy, unstructured community discussions into a structured, searchable beauty intelligence engine. Over 16+ months in production, the site has organically maintained an average Google Search position of 19.7 across competitive cosmetics keywords with almost no traditional backlink building.
Problem Solved
Finding honest beauty product reviews online has become nearly impossible due to sponsored influencers, free-product incentives, and affiliate marketing. While Reddit contains massive volumes of authentic, unvarnished consumer opinions, the discussions are scattered, unstructured, and chaotic to search manually.
Key Features
- Llama 3 LLM classification extracting products, brands, categories, and sentiment
- Entity extraction separating genuine user sentiment from sponsored influencer noise
- Programmatic SEO architecture generating thousands of data-driven, searchable product pages
- Knowledge graph connecting brands, categories, sentiment, and authentic Reddit discussions
- Maintained an average Google Search position of 19.7 across competitive beauty queries with zero backlink campaigns
Architecture Notes
Analytics & Screenshots
What I Learned
- →Good SEO ideas are often good product ideas first: solving real user search intent without manufacture creates organic search traction.
- →AI is far more valuable for structuring messy human conversations than for generating generic SEO articles.
- →Programmatic SEO doesn't have to mean thin content when each page is driven by unique, synthesized community data.
- →Search engines reward deep relational graphs (Product → Brand → Category → Discussions → Sentiment).
Future Improvements
- +Direct CTR optimization on high-impression second-page search queries
- +Skin type and concern filtering (dry, oily, sensitive, acne-prone)
- +Real-time trending alerts when new products surge in Reddit mentions