Sport is experienced through images as well as physical effort: the replay, the diagram, the broadcast and the moment a spectator remembers. In this 2024 research series, I explore how AI might change that visual language. Beyond helping teams analyse performance, it could influence what athletes notice about themselves and how audiences understand a game.

Video and sensor data could be translated into movement paths, comparisons and simulations. A coach might examine how a player finds space or how a technique changes under fatigue. Personalised training could use these views to support a conversation between athlete and coach. The image should make a useful pattern clearer, without reducing a person's performance to a single score.

The same approach could support workload assessment and decisions about recovery. But recognising a pattern associated with injury risk is not the same as predicting an injury in an individual athlete. Visualisations should communicate uncertainty and remain part of a wider assessment. Their apparent precision must not displace medical expertise or the athlete's own account of how they feel.

Tactical analysis might compare formations, reveal recurring situations and explore possible responses during a match. Scouting could draw on a wider range of recordings and competitions. Yet a model trained on established patterns may overlook an unconventional player or reproduce inequalities in the available data. These tools would be most useful when they broaden human judgement rather than define talent on their own.

Officiating requires a distinction: VAR is video assistance for referees, not a synonym for AI. Specific systems, such as semi-automated offside technology, can use AI to support officials. The broader visual question is how to explain a decision. A reconstruction should clarify the evidence and its limits; it should not suggest that every judgement can be settled by an immaculate line on a screen.

For spectators, AI could support personalised highlights, explanatory graphics and alternative viewpoints, including immersive experiences. Someone new to a sport might need context; an experienced viewer might want a tactical reading. More information is not always a better broadcast. The rhythm of the contest and the shared experience of watching should remain visible beneath the analysis.

Predictions could become part of this visual environment, from match probabilities to fantasy sports and betting interfaces. They should be presented as estimates, not promises of results or profit. Athlete data also raises questions about consent, access and commercial use. What helps a team prepare need not automatically become public entertainment.

I imagine sport gaining a richer language for explaining effort, strategy and uncertainty. The opportunity is to help us see more in a movement or a collective decision. Its value will depend on whether these tools deepen our understanding of athletes as people, while preserving the unpredictability that makes sport worth watching.