The final week of the NFL regular season has a handful of teams jockeying for playoff position. Just one upset loss can move a team from hosting a first-round game to an early offseason.

In closely contested contests, a missed call by a referee can make a big difference. Given that artificial intelligence has become ubiquitous in society, and continues to grow, does it have a place on the football field, or more accurately, can it replace referees in getting every call right?

Before answering this question, a review of the rules for NFL football provides a glimpse into the issue.

The NFL Official Rules is a 79-page document that covers the fine detail of how games are played. The document describes 19 specific rules, with detailed diagrams and explanations to help every referee crew get every call right, in every game, on every play.

There are 17 NFL referee crews, each containing nine referees, umpires, judges and officials, including two who handle instant replays. Each goes through extensive training to ensure that the rules get translated into correct calls. Both players and fans rely on these referees to make sure that each game’s outcome is consistent with the play on the field. Given that every ball snap involves 22 players (11 on each team) in motion, referee crews must not only understand all the rules, but they must also be able to apply them in real-time.

Yet AI systems are excellent at learning. Given that the NFL Official Rules book contains well-defined procedures and rule violations, it seems reasonable to conclude that AI should be able to learn the rules and apply them, perhaps even better than the officials on the field. The first conclusion is undoubtedly true. The second is more problematic, given the need to translate these rules into actionable calls immediately.

Some rule violations are easy to detect. When a defensive lineman crosses the line of scrimmage before the ball is snapped, an AI system would be able to identify such an infraction. The issue becomes more challenging when the violation is not a simple quantifiable rule violation but involves a degree of judgment.

For example, defensive pass interference requires not just assessing whether the defensive player inhibits the offensive player from catching the pass, but how they did it. The list of ways that this can occur is long and varied. An AI system may be able to identify many such instances. The question is whether such a system would get the right call more frequently than a referee crew.

What would be ideal is for the NFL to invest in an AI system that watches games and learns how the referee crews call rule infractions. With sufficient time, the AI system would ideally replicate what the referee crews can accomplish.

Given that a referee crew contains nine people, included video review officials, it is highly unlikely that an AI system can match their performance. However, an AI system can serve as a 10th member of the crew, quietly calling the game. After the game, such information can then be compared to what the field crew called, providing a learning exercise not only for the crew, but for the AI system itself.

However, referee crews are more than just a group of people overseeing the NFL rules. When calls are made, fans may support or take offense at the call on the field, adding an extra dimension of intensity to games. Video reviews also provide a brief period of suspense, as fans await the final resolution. In recent years, video reviews have reversed more than half of all calls. This is not an indictment of referee crew skills, but an indicator of how difficult it is to make correct calls.

The likelihood of AI systems replacing NFL referee crews anytime soon is highly remote. The same can be said for NBA and NHL officials. The one sport where AI may have an opportunity to break through is Major League Baseball. Given that baseball involves calling balls and strikes, outs, and balls in play or foul, without players typically interacting, adding a fifth AI umpire is reasonable to expect at some point in the not-too-distant future.

AI systems can learn effectively. Such learning can provide insights and even simulate judgment. Yet sports is a domain where judgment is often the difference between get a call right or wrong. Given how well the current system works, AI will continue to take a back seat when it comes to officiating professional football sports for now. Yet its future role should remain open for discussion.

____

Sheldon H. Jacobson, Ph.D., is a professor in computer science in the Grainger College of Engineering at the University of Illinois Urbana-Champaign. He used his expertise in risk-based analytics to address problems in public policy.

___

©2025 Tribune Content Agency, LLC.