AI Officiating in Sports: What Machines Decide—and What Humans Still Own

Start with the decision, not the AI label

Officiating technology is easiest to understand as four separate layers. A system first senses something, such as ball position, player position or a timestamp. It may then infer an event, compare that event with a rule, and send an alert or call to an official. The last layer is authority: the person or process allowed to change the match record.

Marketing often compresses those layers into a single phrase such as 'AI referee.' That can make a measurement tool sound autonomous and can hide the judgments that remain with officials. A credible description should identify the input, the output, the review path and the owner of the final decision.

• Sense: cameras, ball sensors or timing systems capture match data.
• Infer: software estimates position, contact, timing or another event.
• Apply: the event is compared with a defined rule or threshold.
• Authorize: an official, review centre or automated protocol determines the recorded call.

The cross-sport evidence ledger

The three current examples below sit at different points on that chain. They should not be presented as equivalent proof that AI has replaced officials. Football's semi-automated offside system assists a rule application, basketball's connected-ball trial is collecting research data, and tennis electronic line calling automates a narrow geometric call.

This comparison uses competition and governing-body sources available on 12 October 2026. Those sources establish deployment and operating boundaries. They do not, by themselves, provide independent proof of every performance or accuracy claim.

• Football — input: player and ball tracking; system role: identify positional offside and generate an alert; human boundary: officials still validate the incident and decide interpretive elements such as interference.
• Basketball — input: a Bluetooth-equipped ball in selected NBA preseason games; system role: collect data for possible future officiating uses; human boundary: the NBA says the data does not affect current calls.
• Tennis — input: ball-tracking cameras; system role: produce live 'out' and 'fault' calls at Wimbledon; human boundary: chair umpires still manage the match and the rest of the rules.
• Across all three — evidence still needed: published error analysis by scenario, latency, fallback behavior and a clear correction process.

Football: automation narrows the geometry

FIFA's advanced semi-automated offside technology for the 2026 World Cup combined tracking, kick-point identification, virtual lines and player representations. FIFA said clear positional information could reach on-field officials more quickly. Its test material separates line placement, ball and skeletal tracking, and alert logic, which makes the system easier to evaluate than a single headline accuracy number.

The system does not turn every offside situation into geometry. A player may be in an offside position without committing an offence. Interference with play, interference with an opponent and deliberate play by a defender can require interpretation under the law. The machine can narrow the factual question; the officiating team still owns the football judgment and the fallback when tracking is uncertain.

Basketball: a smart ball is still a research input

On 5 October 2026, the NBA said selected preseason games would use Wilson basketballs containing Bluetooth sensors. The league described the data collection as research for possible future real-time officiating uses and stated that it would not affect calls in those games. Neither the instrumented ball nor the tested referee wrist devices were planned for the regular season.

That boundary matters. A connected object can create new evidence without containing a model, and a future model could advise a review without owning the call. The current public source does not disclose the exact sensor outputs, calibration method, game count, latency or comparative accuracy. The honest description is therefore a controlled data trial, not automated basketball officiating.

Tennis: the narrow call is the product

Wimbledon's official officiating information says live electronic line calling is used on all courts. The task is constrained: determine whether the tracked bounce is in or out and issue the corresponding call. That is closer to direct automation than the advisory chains used for many football incidents, but it still operates inside a larger officiating system.

A line system does not administer the full match. Chair umpires remain responsible for procedure, conduct, score management and other rules. Calibration, camera coverage and a documented contingency still matter because a confident automated call is only as dependable as the sensing and operating chain behind it.

A seven-question evaluation card

Teams, leagues and readers can test an officiating claim without needing access to proprietary code. The right questions expose whether the product is a measurement tool, a recommendation system or an authorized decision-maker—and whether its evidence matches its consequence.

The standard should rise with the stakes. A research-only alert can tolerate more uncertainty than a call that changes possession, a score or a result. Adoption shows that a competition chose to use a system; it does not, on its own, establish accuracy or fairness.

• What physical event does the system observe, and what does it infer?
• What rule or threshold converts the measurement into an alert or call?
• Which official can inspect, reject or correct the output?
• How was the full matchday chain tested—not only the model in isolation?
• Are misses, false alerts and latency reported by scenario?
• What happens when cameras, sensors, calibration or communications fail?
• Can participants and viewers understand who made the final decision?

Limitations and source notes

This guide compares public implementations in three sports and does not claim to catalogue every officiating system. Baseball strike-zone systems, cricket decision review, rugby ball tracking and judged-sport scoring require separate rule and validation analysis. Some technologies use computer vision or statistical models; others are better described as sensors, geometry or automation. We do not assume every automated system is AI.

FIFA, the NBA and Wimbledon sources establish what each organization says is deployed or being tested. They are authoritative for operating status but are not independent audits of performance. The peer-reviewed exploratory study is included for its wider human-and-system framing, not as validation of any named product. No unpublished data, vendor demonstration or medical claim is used. Sources were checked on 12 October 2026.