Building an AI-Driven Content Ideation System

Marketing Against the Grain2026-07-30go watch the original →

the gist

Kieran Flanagan outlines a content ideation system that uses a custom audience profile and a dataset of 160+ high-performing posts to generate platform-specific content ideas, ensuring output remains aligned with personal brand voice rather than generic AI slop.

The Content Ideation Framework

The author rejects the standard "prompt-to-post" workflow, which often results in generic, low-quality content. Instead, he uses a structured system that functions as a persistent "skill" within Claude and ChatGPT. This system relies on three core inputs: a rich audience profile, a database of proven platform-specific content patterns, and live trend research via MCP servers. By maintaining shared context between AI assistants, the system ensures that content ideas are consistently mapped to the creator's specific audience and platform requirements.

Audience Profiling and Pattern Mapping

The system distinguishes between an Ideal Customer Profile (ICP) and an "audience profile." The latter focuses on emotional triggers, trusted voices, and situational frames rather than firmographics. The author maintains a dataset of 160+ high-performing posts from the past year, categorized by platform (LinkedIn, Substack, YouTube). When generating ideas, the AI references this dataset to identify patterns—such as "contrarian takes," "data nuggets," or "identity reckonings"—rather than just topics. This allows the creator to receive ideas that are statistically likely to perform well based on historical engagement data.

System Maintenance and Refinement

The author treats the AI as a research and brainstorming engine rather than a writer. He manually curates a "content queue" by filtering AI-generated suggestions, removing those that do not fit his current focus, and adding new ones based on live trend research from Reddit, X, and the web. He emphasizes that the creator must remain the "AI strategist," using the tool to build the foundation of a post while applying personal taste and judgment to the final execution. The system is updated monthly with new performance data to prevent the ideation logic from going stale.