As large language models (LLMs) increasingly mediate how users discover brands, keyword research alone can no longer map the full landscape of user intent. This talk presents a real case study involving a global peripherals brand, showing how a structured five-stage framework transformed client data and research inputs—including brand guidelines, Google Search Console (GSC) data, competitor gaps, and People Also Ask (PAA) questions—into over 1,000 semantically categorised prompts, organised by funnel stage, sentiment, and persona.

The framework used an AI model to build a variable dictionary, generate reusable stem structures, and expand them into natural language prompts that mirror real user behaviour across the web, Reddit, and YouTube. The result was a content strategy grounded in how LLMs actually synthesise evidence, along with measurable growth in AI citation coverage.