AI in Sports Officiating: How Automated Decision Support Is Changing Refereeing

These tools can help determine whether a ball crossed a line, whether a player was offside, where a pitch passed through the strike zone, or whether a gymnastics element met defined technical requirements. Their value is greatest when a rule can be translated into observable measurements and the system has been tested under realistic competition conditions.

The International Olympic Committee’sOlympic AI Agendaidentifies judging and refereeing as one of the areas in which AI may support sport. At the same time, current systems demonstrate why human oversight, transparent rules, technical certification, and procedures for handling errors remain essential.

This guide explains how AI-assisted officiating works, where it is already being used, which benefits it may provide, and what sports organizations should consider before relying on automated decisions.

What Is AI-Assisted Sports Officiating?

AI-assisted officiating uses computational systems to help referees, judges, or umpires observe, classify, measure, or review events during a competition.

The technology may involve:

Not every electronic officiating system is technically an AI system. A photo-finish camera, for example, records an image that a human official interprets. Goal-line technology may calculate whether the ball crossed a defined plane without making subjective judgments about fouls or player intent.

The distinction matters because different systems require different forms of governance.

A measurement system answers a narrowly defined question, such as whether the whole ball crossed a line. A decision-support system may identify evidence and recommend a result. A more autonomous system may classify an action and issue a decision with limited human involvement.

Most established sports technologies remain somewhere between measurement and human-supervised decision support.

Why AI Is Becoming Important in Officiating

Competition is faster than human perception

Elite athletes and balls can move too quickly for an official to observe every relevant detail from one position.

A referee may also need to monitor contact, player location, timing, possession, and rule context simultaneously. Cameras and tracking systems can examine several angles or data points that are not available to one human observer.

Small decisions can have major consequences

An offside position measured in centimeters, a tennis ball landing near a line, or a pitch passing close to the edge of the strike zone can influence a match result.

As broadcasts provide viewers with slow-motion replays and digital graphics, incorrect calls also become highly visible. This increases pressure on sports organizations to provide consistent and explainable decisions.

Human judgment is affected by context

Referees develop expertise through training and experience, but judgment can still be affected by viewing angle, obstruction, fatigue, time pressure, crowd noise, and the complexity of the incident.

A study of video assistant refereeing in football found that video review improved decision accuracy in the situations examined, although reviews also affected the time required to reach a decision.

Technology may reduce some perceptual errors, but it does not automatically eliminate disagreement. Many rules still require interpretation of intent, severity, advantage, or whether contact had a meaningful effect on play.

Tracking infrastructure is improving

Modern stadiums increasingly contain synchronized cameras, connected balls, communication networks, replay systems, and real-time data platforms.

Once this infrastructure is available, the same data may support officiating, broadcasts, performance analysis, and fan visualizations. This can make investment more commercially attractive, although combining these uses raises questions about access, ownership, privacy, and system independence.

How AI Officiating Systems Work

A typical AI-assisted system contains several connected stages.

1. Data capture

The system gathers information from cameras, sensors, microphones, or connected equipment.

Camera placement, frame rate, lighting, image resolution, network stability, and synchronization all influence the quality of the evidence. A model cannot reliably identify an event that the equipment did not capture accurately.

2. Detection and tracking

Computer vision identifies relevant objects or body positions.

Depending on the sport, this may include:

Tracking software then estimates how these objects move over time.

3. Rule-based calculation or classification

The system applies a defined rule or model.

For a line decision, it may calculate the relationship between the ball and a boundary. For an offside decision, it may identify the relevant body parts, the second-last defender, and the moment the ball was played.

In judged sports, the process can be more complex. A pose-recognition system may compare an athlete’s movement with technical definitions in the sport’s code of points.

4. Confidence and exception handling

A responsible system should identify when the evidence is incomplete or uncertain.

Possible problems include:

Uncertain cases should be escalated rather than forced into an apparently precise decision.

5. Human review or automatic output

Some systems immediately issue a call. Others notify an official, who reviews an image or animation before making the final decision.

The appropriate level of automation depends on the clarity of the rule and the consequences of error.

Current Examples of Technology-Assisted Officiating

Sport and system Technology role Human role Main decision

Football semi-automated offside Tracks players, limbs, ball position, and timing Video officials validate the event and referee applies the decision Offside position

Football VAR Provides replay evidence for defined match-changing incidents Referee retains the final decision Goals, penalties, red cards, mistaken identity

Tennis electronic line calling Tracks ball trajectory and bounce location Depends on live-call or challenge format In or out

Baseball ABS Challenge System Tracks each pitch relative to the batter’s defined zone Umpire makes the first call; eligible players may challenge Ball or strike

Gymnastics Judging Support System Uses 3D sensing and element-recognition technology Judges use the output as support Recognition of technical elements

Taekwondo experimental AI review Analyzes video and body movement Researchers recommend referee confirmation for difficult cases Penalties and review decisions

Football: From Video Review to Semi-Automated Offside

Football demonstrates the difference between replay technology and AI-assisted event detection.

Under theIFAB VAR protocol, the video assistant referee may intervene only in specified categories of match-changing incidents. The referee must make an initial decision, and the final decision remains with the referee. The protocol is intended to correct a clear and obvious error or serious missed incident rather than re-officiate every moment of the match.

VAR itself is primarily a structured video-review process. Its operation may use advanced replay and tracking tools, but the referee still interprets the Laws of the Game.

Semi-automated offside technology is more computational. FIFA’s system combines player and ball tracking with automated identification of possible offside situations. The technology can help determine the relevant kick point and the positions of players before presenting the evidence to video match officials. FIFA has continued developing the system for the 2026 World Cup, including more advanced skeletal and ball-tracking data. (FIFA — Semi-Automated Offside Technology)

The word “semi-automated” is important. The system accelerates detection and measurement, but match officials remain responsible for validating the relevant players, interpreting active involvement, and communicating the final decision.

Tennis: Electronic Line Calling

Tennis line decisions are particularly suitable for automation because the central question is narrow: did the ball touch the court inside or outside the relevant boundary?

Electronic line-calling systems reconstruct the ball’s trajectory using synchronized cameras or other approved tracking methods. Some systems are used only when a player requests a review. Others provide live line calls for every bounce.

TheInternational Tennis Federationevaluates systems for accuracy, reliability, practicality, and suitability. Since 2025, classified systems can receive Gold, Silver, or Bronze status. Approval is specific to factors such as surface, setting, and whether the system provides challenge-based or real-time calls. Poor continuing performance can lead to suspension of classification.

This model illustrates an important governance principle: a provider should not receive general approval simply because its system performed well once. Certification should reflect the environment in which the technology will actually operate.

A system validated on an indoor hard court may not automatically be appropriate for clay, grass, outdoor light, or a lower-cost camera setup.

Baseball: Human Calls With Automated Challenges

Major League Baseball introduced its Automated Ball-Strike Challenge System for the 2026 season after testing in the Minor Leagues and Spring Training.

The home-plate umpire continues to call every pitch. A batter, pitcher, or catcher may immediately challenge a ball-strike decision. The system then tracks the pitch relative to the batter’s defined strike zone and displays whether the original call is confirmed or overturned. (MLB — ABS Challenge System)

This hybrid model preserves the traditional role of the umpire while providing a limited correction mechanism. It also makes technology use part of game strategy because teams have a restricted number of unsuccessful challenges.

MLB reported that during 288 Spring Training games in 2025, there were an average of 4.1 challenges per game, with each challenge taking an average of 13.8 seconds. These figures are specific to that testing period and should not be assumed to represent every future season.

The system also reveals that automation does not remove rule design. The league must define where the strike zone is measured, how it changes with batter height, who may challenge, how quickly the request must be made, and what happens if technology is unavailable.

Gymnastics: AI as a Support for Technical Judging

Artistic gymnastics involves complex movements that can be difficult to evaluate consistently from a fixed viewing position.

World Gymnastics and Fujitsu have worked on a Judging Support System that uses 3D sensing and body-position recognition to identify elements and provide additional visual evidence. The collaboration began in 2017, and the system has been used at official competitions since 2019. (World Gymnastics — Judging Support System)

This type of application differs from line calling. A gymnastics routine may involve questions about element identity, body angle, rotation, execution, connection, and artistic requirements.

AI can help measure body positions or recognize patterns, but a complete score may still contain evaluative judgments that are difficult to reduce to one measurement. World Gymnastics presents the technology as judging support rather than a universal replacement for the judging panel.

The system may also be valuable for education. Coaches and judges can review how movements were recognized, which may support more consistent interpretation of technical rules.

What Research Says About AI Review

Research into fully or partly automated sports judging remains limited and varies considerably by sport.

A 2025 feasibility study reanalyzed video reviews from the Paris Olympic taekwondo competition using AI tools. The researchers reported improved efficiency and substantial consistency with expert decisions, but they also concluded that human oversight remained important for ambiguous or complex cases. They proposed a hybrid process in which AI conducts initial analysis and a referee confirms the result.

This is a useful general principle, but the results should not be generalized automatically to other sports or competitions. The study used existing broadcast video, a defined sample of reviewed incidents, and particular AI tools. A live production system would require additional validation, integration, reliability testing, and regulatory approval.

A broader systematic review of ethical issues in sports AI found that topics such as bias, privacy, transparency, accountability, and athlete rights remain underdeveloped compared with the pace of technical adoption.

Potential Benefits

More consistent measurement

Technology can apply the same geometric or timing rule repeatedly without fatigue.

This is especially useful for binary or narrowly defined decisions such as line calls, finishing order, or whether a tracked object crossed a boundary.

Access to evidence unavailable to one official

Multi-camera tracking can reconstruct an event from several viewpoints. Sensors may capture timing or contact that cannot be reliably observed from the field.

Faster review

Automated event detection can direct officials to the relevant moment instead of requiring them to search through a long sequence manually.

Faster decisions may reduce disruption, although speed should not come at the expense of accuracy or procedural fairness.

Improved training

Stored incidents, three-dimensional reconstructions, and system outputs can help officials compare decisions and study difficult situations.

Technology may therefore support referee development even when it is not used to make live calls.

Greater transparency

Animations and replays may help athletes and spectators understand why a decision was made.

However, a polished graphic is not proof that the underlying system was correct. Transparency also requires information about rules, measurement uncertainty, review procedures, and system limitations.

Limitations and Risks

AI can produce false precision

A graphic may show a sharp line or exact coordinate even though the camera image, sensor data, or model contains uncertainty.

Sports organizations should avoid presenting estimated positions as perfectly measured facts.

Subjective rules remain subjective

AI is more suitable for locating a ball than deciding whether a tackle was reckless, whether contact was intentional, or whether an artistic performance met a qualitative standard.

Models can classify past examples, but historical decisions may themselves contain inconsistency or bias.

Training data may be unrepresentative

A computer-vision model trained on one venue, uniform style, body type, camera arrangement, or lighting condition may perform differently elsewhere.

Testing should include different athletes, competition levels, equipment, weather conditions, and uncommon incidents.

Automation bias can influence officials

Officials may give excessive weight to a system recommendation because it appears objective.

Human review is meaningful only when officials are trained to challenge the technology and have access to enough evidence to do so.

Technology may change the sport

An automated system does not merely enforce an existing rule. It can change player behavior, tactics, game flow, and expectations.

The MLB challenge format, for example, introduces strategic decisions about when to challenge. Football video review can affect celebrations and restart timing.

Access is unequal

Elite competitions can afford camera arrays, backup systems, certified technicians, and dedicated review rooms. Community and lower-level competitions often cannot.

If different levels of the same sport use different decision systems, athletes may experience inconsistent rule enforcement and development pathways.

System failure must be anticipated

Networks can fail, cameras can be obstructed, sensors can disconnect, and software can produce invalid output.

The rules must specify whether play continues with human officiating, whether earlier decisions remain valid, and who declares the system unavailable.

Practical Implementation Framework

1. Define the exact decision

Do not begin with the objective of “using AI.”

Define the question the technology must answer:

Narrow decisions are usually easier to validate than complete automated refereeing.

2. Preserve a clear rule basis

The output must correspond to an official rule.

If the rule contains ambiguity, the governing body should clarify that ambiguity rather than allowing the vendor’s model to define the sport informally.

3. Test under competition conditions

Testing should include:

The ITF classification process and FIFA Quality Programme provide examples of sport-specific technical evaluation rather than relying only on manufacturer claims.

4. Measure the right outcomes

Evaluation should cover more than overall accuracy.

Relevant indicators may include:

5. Define human authority

The rules should state:

6. Train officials before deployment

Officials need operational and critical training.

They should understand what the system measures, what it does not measure, how errors appear, and how to continue when technology fails.

7. Communicate with athletes and audiences

Participants should know how the system affects the competition.

Public explanations should avoid technical jargon while accurately describing the decision process. A simple animation should be accompanied by clear rules and consistent procedures.

8. Audit performance continuously

Approval should not be permanent.

Software updates, new equipment, venue changes, and revised rules can alter performance. Systems should be retested, and significant errors should be documented and investigated.

Will AI Replace Sports Referees?

Full replacement is unlikely across most sports in the near future.

Many decisions combine measurement with interpretation, communication, conflict management, safety oversight, advantage, and understanding of match context. A referee also controls restarts, manages participants, responds to unexpected events, and applies rules in situations that may not appear in a technical dataset.

Automation is more likely to expand in layers:

Hybrid systems may provide the most practical balance. They allow machines to perform repetitive measurement while preserving human responsibility for context-sensitive decisions.

Future Outlook

The next generation of officiating systems will likely combine higher-resolution cameras, connected equipment, real-time pose estimation, edge computing, and automated event recognition.

Football systems may detect more restart and boundary events. Gymnastics and other judged sports may receive more detailed body-angle and element-recognition support. Combat sports may experiment with automated identification of contact, prohibited actions, and scoring events.

Centralized remote officiating may also grow. World Rugby announced a remote Television Match Official hub in 2026 as part of an effort to improve consistency and reduce the infrastructure required from individual competition organizers. This is not fully automated AI judging, but it demonstrates how technology can reorganize officiating operations as well as individual decisions. (World Rugby — Remote TMO Hub)

The central policy question will not simply be whether AI is more accurate than a referee. Sports bodies will need to decide which form of accuracy matters, which errors are acceptable, how uncertainty is communicated, and who is accountable when the system fails.

Mixed FAQ

Is VAR an artificial intelligence system?

VAR is primarily a structured video-review process involving human match officials. AI and automated tracking may support particular functions, such as offside detection, but VAR itself is not an autonomous AI referee.

What is semi-automated offside technology?

It uses player and ball-tracking data to identify possible offside situations and generate visual evidence. Video match officials validate the information before the decision is applied.

Can electronic line calling make mistakes?

Yes. Performance can be affected by calibration, camera coverage, surface conditions, hardware failure, or unusual events. Certified systems are tested, but no technology should be treated as infallible.

Why not automate every ball and strike in baseball?

MLB chose a challenge model that preserves the umpire’s initial call while allowing players to correct selected decisions. This balances automation with human officiating and adds a strategic element.

Can AI judge artistic sports fairly?

AI can measure body positions and identify defined elements, but artistic and execution scores may include context and qualitative judgment. Human-supervised support is currently more realistic than complete automation.

Who is responsible when an automated decision is wrong?

Responsibility should be defined by the governing body before deployment. It may involve competition officials, technology providers, system operators, and the organization that approved the system.

Does AI guarantee unbiased officiating?

No. AI can apply criteria consistently, but its models, data, thresholds, and rules are created by people. Bias can enter through training data, system design, validation, or unequal performance conditions.

Sources