Survey Maps the Landscape of AI-Generated Sports Commentary

A new survey published on arXiv examines the emerging field of AI-Generated Game Commentary (AI-GGC), a research area that has expanded as artificial intelligence capabilities have matured. The work offers a systematic organizing structure for a field that has grown somewhat scattered across different approaches and applications.

Why AI Commentary Matters#

As AI systems become more capable, they can produce sports commentary and game narration at scale without the constraints of human broadcasters. This technology promises advantages like round-the-clock availability, tailored narration for different audiences, and coverage of events that might not otherwise receive professional commentary. However, researchers note that studies in this space have largely developed in isolation, lacking a shared conceptual framework.

A Three-Part Model for Commentary#

The researchers propose a taxonomy built on three core capabilities that any commentary system must demonstrate:

• Live Observation — perceiving what is happening in real time across multiple types of sensory input
• Strategic Analysis — understanding the tactical implications and significance of unfolding events
• Historical Recall — drawing on context, past performances, and background knowledge to enrich narration

These capabilities map to three corresponding commentary styles: descriptive commentary (what is happening), analytical commentary (why it matters strategically), and background commentary (contextual information from prior events and statistics).

Next Steps for the Field#

The survey reviews existing methods, datasets used to train and evaluate these systems, and metrics for assessing performance. The authors identify limitations in current approaches and outline open questions for future work, suggesting where the field should focus effort to make AI commentary systems more robust and useful.

View the full paper (PDF)