The Moment No Model Saw Coming: How AI Changed the Way We Watched the 2026 World Cup

Late in the first half of England’s quarterfinal against Norway, Norway’s goalkeeper launched a goal kick that clipped one of the cables holding up the camera that flies over the pitch. The ball dropped, England picked it up, and a few passes later, Jude Bellingham scored the equalizer. Norway protested, the referee hadn’t seen it, and VAR didn’t call.

More than an hour after the match, FIFA settled the argument by consulting the ball. A sensor inside it tracks what FIFA calls the ball’s “heartbeat,” and according to FIFA’s statement, it showed no spike while the ball was in the air, so there was no evidence of contact and the goal stood. I keep coming back to that goal because it sums up what I saw this summer. There was technology in the ball, on the referee and all over the broadcast, and I wondered if fans gave it a second thought.

This is the last post in our series on AI and machine learning at the 2026 World Cup. Since the first one, we’ve looked at predicting the winner, scouting, injuries, VAR and ticket prices. This one is about the fans.

I sat down one more time with Finn McCallum, the player and coach who has been my guide to the soccer side of this series. He watched games on his phone, at home, with friends, in the stadium in Seattle and in a bar at the bottom of the ski lifts in Whistler, so he saw the tournament just about every way a fan could.

More Cameras, and a Camera on the Referee

The first thing Finn noticed was how many more cameras there were, and how much sharper the picture looked from above and around the field. Not all of them were there for the broadcast. Some existed so the officials reviewing a play didn’t have to scroll through dozens of feeds to find the one angle that showed what happened.

The new angle he liked best was the referee’s own. FIFA and Lenovo built a new version of Referee View, a body camera whose footage gets smoothed by AI stabilization software in real time so it doesn’t look like a shaky phone video. Finn thought it helped players and fans see things from the referee’s side. If you’re the player who thinks you got fouled, you only ever see the play from your own eyes. Then you watch the referee’s angle and realize he barely touched you. Even when the referee gets it wrong, Finn said, you can at least understand where he was coming from.

The same sensor that cleared England’s goal had already ended Croatia’s tournament a week earlier. Deep in stoppage time of their round of 32 match against Portugal, Croatia scored what looked like the equalizer. The chip in the ball registered a touch off Igor Matanović’s head, which put a teammate offside, and the goal was taken off the board. Matanović admitted afterward that he felt slight contact with his hair. Finn wasn’t convinced the ball touched him at all, and plenty of people watching on Fox felt the same way.

3D Avatars and the Offside Call

The biggest visual change for fans was in the offside replays. Before the tournament, Lenovo scanned every player (each scan takes about a second) and built a 3D avatar matching that player’s actual body dimensions. When semi-automated offside technology makes a call, the replay now shows the real players instead of the generic mannequins from 2022.

Finn walked me through why offside is so hard for humans in the first place. The assistant referee has to judge a line across the whole width of the field, usually with a defender or an attacker blocking the view. A hand ahead of the line doesn’t count, since you can’t legally play the ball with it, but a shoulder does. Officials have also been told to keep the flag down until the play finishes, so now you’ll see a player score, celebrate, and then watch the flag go up for something that happened 15 or 20 seconds earlier.

The avatars make the replay easier to follow, since they show what the cameras and the ball measured on a body that looks like the actual player. A clearer picture didn’t help Croatia, though. The call was technically correct and still felt wrong to most people watching, which is the same problem I wrote about in the VAR post.

The Broadcast Is Competing for Your Attention

I have a theory about all those cameras, and Finn thought it was plausible. Soccer now fights TikTok, YouTube, streaming and video games, along with every other sport, for a slice of people’s attention, especially people younger than Finn. If FIFA wants to stay on everyone’s screens, the broadcast has to look great whether you’re in the stadium, at a bar or on a phone on the bus.

One of FIFA’s bigger moves was a deal with a streamer. IShowSpeed, one of the most popular creators in the world, signed with FIFA, Fox and YouTube to stream matches from host stadiums and his home studio alongside live match footage, which no creator had been allowed to do before. Finn saw it as FIFA using streaming culture to reach younger fans instead of competing with it.

The extra cameras also make it easier to cut a match into short clips, and Finn’s feed was full of them, from Fox’s official accounts and from random TikTok accounts posting whatever moment they liked. FIFA’s own numbers show how much of the tournament lived on phones. By the end of June, it reported 30 million unique visitors to its World Cup app, up 130% from 2022, and 16 million new TikTok followers during the group stage alone.

The broadcast also made a point of showing which celebrities were in the building. Finn said he noticed every time, and he still announced their names to whoever he was watching with, because that’s what everyone does. A famous face in the stands tells viewers this is a big cultural event. He didn’t think the cutaways annoyed fans, since directors usually waited until a team was passing it around in its own half or the ball was out for a throw-in. Not every cameo went over well. At the USA’s win over Australia in Seattle, Finn was in the stadium when Health Secretary Robert F. Kennedy Jr. appeared on the big screen, and he says the crowd booed loudly enough to be heard through the field microphones.

Apps, Fan Zones and Sports Betting

Finn didn’t know the FIFA app had live tactical data until he researched it for our interview. It had real-time expected goals, momentum graphs and even a camera that followed a single player. FIFA’s Play Zone, its hub for free games, reached more than 33 million fans over the tournament.

He noticed the gambling a lot more. Sports betting apps were all over the ad breaks, and Finn’s point was that betting gets people to watch games they would otherwise skip. His example was Algeria against Jordan. He doubted many American viewers could find either country on a map, but the game, a 2-1 comeback win for Algeria, still drew people who had money on it.

The watch parties were the part he remembered most. In 2022 he was in school during most matches, and in 2018 he was too young to head into Seattle on his own. He doesn’t remember anything like the FIFA-organized fan zones this time, which drew nearly 2 million visitors in the first round of matches alone. As I covered in the ticket pricing post, plenty of fans were priced out of the stadiums, and for them the fan zones, bars and phones were the World Cup. Even friends of his who barely care about soccer were watching games and asking him questions. He admitted that wasn’t always welcome, since he was trying to watch.

The brand partnerships mostly worked on him too. His favorite was Michelob Ultra’s “Tab Closed,” which turned a bar receipt into a tally of Lionel Messi’s career across his World Cups. He liked it enough to consider it for a marketing project at school. Finn’s view was that the celebrities in the stands end up working as ads for the World Cup itself.

Most People Don’t Know AI Is There

When I asked whether his friends outside the tech world notice any of this, Finn said most people watching sports have no idea AI is involved beyond VAR. They don’t care much either, as long as it doesn’t take away from the game.

The friends who do care tend to dislike AI. Most of them are studying for careers in healthcare, and Finn said their objection is less about AI itself and more about the water and energy that data centers consume. One friend turned off Siri on her new iPhone because of it. Finn uses AI regularly, to help with assignments rather than do them, and he figures anyone entering the workforce without at least a basic understanding of it will fall behind.

In my experience, people either love AI or want nothing to do with it, and I rarely meet anyone in between. I’m closer to Finn than to his friends. If I had cancer, I’d want a computer vision model reading my scans, because it can catch things a person might miss. His friends have a point about the data centers, though, and I don’t have a good answer for them yet. At this World Cup, most fans seemed fine with the technology as long as it stayed in the background.

Eight Minutes in Whistler

For the end of this series, I asked Finn for a moment from the tournament that no algorithm could have predicted. He picked Cape Verde. Cape Verde, a country of about half a million people playing in its first World Cup, didn’t win a single group match. It didn’t lose one either. Draws with Spain, Uruguay and Saudi Arabia made it the smallest country ever to reach the knockout round, which set up a round of 32 match against the reigning champion, Argentina, in Miami.

Everyone expected a blowout, Finn included. Messi scored early. Cape Verde equalized in the second half, and for long stretches it was the better team. Argentina went ahead in extra time, and that should have been the end of it. Then in the 103rd minute, Sidny Lopes Cabral curled a shot into the top corner to make it 2-2.

Finn watched that goal at the bar in Whistler. The screen faced a crowd on the patio, and a second screen on its back faced another crowd out by the lifts, which run mountain bikers in the summer. When Cabral scored, both sides went up at once. People of every age were jumping, and so were the servers. Finn said he’d never seen that many strangers that happy at the same time.

Argentina won it eight minutes later on an own goal and went on to the final. For those eight minutes, though, Cape Verde was level with the world champion, and Finn has a hard time believing any simulation, run as many times as you like, would have produced that game.

That takes me back to the prediction post. Before the tournament, the leading model gave Spain about a 16% chance of winning. Spain beat Argentina 1-0 in extra time in the final, and Finn had picked them from the start. That result doesn’t contradict the model, which only ever said the favorite loses more often than it wins. Spain needed extra time to win the final, and Cape Verde came within eight minutes of taking Argentina to penalties.

When I started this series in June, the final felt a long way off. By the time Spain lifted the trophy in New Jersey, I was a little sad to see it go, the same feeling I get at the end of a summer I spent months planning around.

The series opened with two ways AI enters the game: on the pitch, helping teams win, and off the pitch, shaping what fans see and pay. Most of these posts stayed on the pitch, but off the pitch is where most fans actually ran into the technology, usually without knowing it.

Thank you to Finn for every interview, tangents included. A lot of my favorite material came from those.