Generative Engine Optimization Pipeline
The problem
A CDMO client needed to know whether their brand surfaces in AI-model answers to real buyer questions, and which competitors appear in their place. A new form of SEO for the LLM era.
What I built
A production Python pipeline that queries four frontier models (OpenAI, Anthropic, Google, Perplexity) through OpenRouter and writes structured, analyst-ready results to Excel. A 57-prompt research library across query clusters (Discovery, Comparison, Selection, Thought Leadership), with priority-keyword coverage and customer size and stage variants to stress-test visibility against 16 named competitors.
The outcome
Production reliability: response validation rejects silent truncations, automatic retries absorb transient model errors, and every run writes a timestamped backup before any changes. The output is analyst-ready data for monthly brand-visibility tracking.
Source
Private. Code and prompt library available on request under NDA.
Interested in this kind of work? Get in touch