Tech stack: Built on Next.js/React, pulling data from the Ahrefs API and Open Food Facts API.
We used OpenAI's API for the heavy lifting - deduplicating, aggregating, and categorizing 800,000+ products to build comprehensive profiles across 609 food additives.
The data work was surprisingly tricky. We had to refine our Ahrefs methodology to use "matching terms" rather than "related terms" - otherwise we got flooded with irrelevant keywords (MSG search volume was massively inflated by "Madison Square Garden" ). Filtered for 100+ monthly searches to eliminate spam while keeping meaningful queries.
The result: real search demand data paired with actual product usage patterns, so you can see both what people are curious about AND where additives actually show up in products.
Happy to answer any technical questions!
All the work was done manually. We just selected those essays which has a structure, summarized either in bullet points or headings.
We edited each fact, reducing text, because with icon you don't need as much text.
Regarding automation I experimented with text extraction (using abstractive text summarization) but haven't got good results yet.