AI Schema.org JSON-LD Module Introduces Prompt-Based Structured Data in Drupal

Drupal contributor Jacob Rockowitz has introduced an AI Schema.org JSON-LD module that uses prompt-based workflows to generate structured data for Drupal content entities.

The module combines Drupal AI ecosystem tools including AI Automators, Field Widget Actions, and JSON Field to generate Schema.org JSON-LD from entity content and contextual prompts. Content and instructions are sent to an AI provider, which returns machine-readable structured data that can be reviewed, saved, and embedded into webpages. A supporting sub-module also allows prompts and responses to be logged for iterative refinement.

The implementation supports multiple entity types, including nodes, media, and taxonomy terms, while storing structured data as native JSON. According to Rockowitz, this approach enables the creation of a queryable knowledge graph and allows AI systems to interpret complex nested content structures such as paragraphs and components.

The module is positioned as an alternative to traditional Schema.org workflows that rely on manual configuration. Instead, it centres on prompt-and-response interactions, where prompts are refined through testing across AI models and Schema.org references. The approach also reflects broader changes in how structured data is produced and consumed, particularly as AI systems increasingly use Schema.org markup for information retrieval.

Rockowitz states that the module was built using AI coding agents with human oversight and encourages contributors to maintain test coverage and improve prompt strategies. He also frames the project as part of a wider transition in which AI systems both consume and generate structured data, requiring developers and site owners to adapt their workflows while retaining human review processes.