From Descriptive to Behavioural Metadata
Traditional metadata describes content: its title, author, publication date, format, and perhaps a few subject tags. This metadata serves findability — helping users and search systems locate content that matches a query. It is designed for retrieval by keyword proximity.
Behavioural metadata describes how content should be used: which audience segments it serves, at which stage of which journey, for which intent, in which context. It enables AI systems to select content for specific purposes — not just retrieve content that matches a search, but assemble the right content for the right person at the right moment.
The Metadata Stack
Layer 1 — Descriptive metadata: Title, author, date, format, word count. The minimum viable metadata set for content management. Layer 2 — Structural metadata: Content type, template, component identifiers. Enables structured content assembly and reuse. Layer 3 — Behavioural metadata: Audience segment, journey stage, intent classification, channel suitability. Enables personalisation and AI-driven content selection. Layer 4 — Semantic metadata: Entity references, relationship tags, concept classifications. Enables reasoning about content meaning and context.
Key Takeaways
1. Metadata is no longer primarily a findability tool — it is the operational fuel for personalisation, AI retrieval, and recommendation logic.
2. The metadata stack has four layers — descriptive, structural, behavioural, and semantic — each serving different AI system needs.
3. Metadata governance — the process for maintaining accuracy, completeness, and consistency — is what determines whether the metadata investment retains its value over time.