Long, unstructured documents can be difficult to find, manage, and work with for both people and AI. A clear structure is essential for transforming and dividing disorganized content into smaller, usable, and more valuable information assets. A well-formed content structure enables componentization of content and the ability to add content to a knowledge graph. It also facilitates efficient reuse, personalization, and discoverability across platforms and contexts.

At Text Analytics Forum 2025, Joe Hilger and Kyle Garcia of Enterprise Knowledge discussed how combining large language models (LLMs) with content models enables LLMs to reference structured blueprints that define components and their required elements. By breaking content into well-defined parts, this approach improves consistency, enhances reusability, and makes content easier to manage and scale across platforms.

Participants in this session learned:

• How LLMs parse through documents and how long, unstructured documents hinder this process;
• The value of providing an LLM concise, contextual content elements to serve as a source of truth;
• The key components of a content model and the utility a content structure provides, and;
• How structured information in a content graph can support analytics, power recommendation and search, and augment an LLM’s capabilities.