Every game, everywhere, all the time: Inside FOX Sports' winning AI strategy for World Cup '26

Major upsets, heroic goals, showstopping saves, and new records by some of the world’s greatest players of all time — this summer’s World Cup far exceeded expectations, which were already huge given the historic expansion of the tournament to 48 teams and 104 matches.

For the team at FOX Sports, the ambitions were just as big as the games themselves. So was the pressure.

More people were watching in more ways on more platforms than ever before. For the final, 51 million-plus tuned in at some point, and five games surpassed 20 million viewers; across FOX digital and social channels, there were a staggering 17 billion content views across the entire tournament. And that’s to say nothing of all the pickup the broadcast got on third-party platforms. 

With so many eyes on them, FOX Sports had to deliver highlights with the flair of France, the resilience of England, the passion of Argentina, and the technical brilliance of Spain.

Fortunately, AI has come off the sidelines and is now firmly embedded in the editing suite.

“We’re trying to get to a simple place,” Dustin Myers, FOX Sports’ executive vice president for production operations and technology, said in a recent interview, “where all 1,200 people here working on this huge thing that is the World Cup can find everything. And I mean everything, every highlight, soundbite, and clip. And they can find it from anywhere in the world in a matter of seconds.”

Now, all it takes is a simple search, like you might do on YouTube or other platforms: “ overhead and ref-cam view of Messi’s three goals against Algeria;” “all shots of former teammates Jude Bellingham and Erling Haaland embracing during England-Norway match;” “every time Zlatan smirks during commentary,” for a funny social reel. 

To most modern web users, such functionality might feel utterly unremarkable, but that’s rarely been the case for the broadcast industry.

“You go to these trade shows, and everything is kinda clunky, there’s still all this analog thinking in a digital, and really now conversational-technology age,” David Sobel, vice president for media management at FOX Sports, said. “We all have these great consumer technology experiences, and that’s what we’re trying to bring to our team, so they can do their best work during these big events, as well as day to day.”

At the same time, these new tools can in no way jeopardize a broadcast with tens of millions of fans tuned in, and millions more waiting for highlights and social content to appear within minutes of the final whistle. We recently sat down with Myers and Sobel to learn more about the technical demands their teams faced during the biggest World Cup ever, how AI helped tackle them without interrupting the action, and what’s the next stage for tech in sports broadcasting.

How do you prepare for an event that’s already the biggest in the world, and suddenly it’s 60% bigger, in terms of total matches?

Myers: Even in The States, this was already the hardest event in the pyramid of sports. For broadcasters, we’ve got stuff like soccer and hockey, then basketball, baseball, then of course football, and above that’s the Super Bowl. And then it’s the World Cups and Olympics. 

It’s not just that they’re bigger events, but it’s day after day, like producing multiple primetime games every day for weeks on end. There were 96 games before we even had a day off, 50% more than the entire last tournament. It’s unprecedented, and every one of those matches has to be bulletproof — flawless not only live but on all the other platforms we’re serving. 

Sobel: This is where we start to look for how technology can make a difference, not only in delivering new experiences but relieving some of the burden on the team, too. We approached this knowing you couldn't do it the way you've done past events.

AI tools helped FOX Sports accelerate production across post-game highlight reels as well as more narrative digital features and social-first content.

Can you give us some examples of that new approach?

Myers: For one thing, this is the first time we’ve ever had a World Cup or Olympics across three countries, with 16 host cities. In the past, you’d pick a place, like Paris or Doha, and that would be your hub and you’d go out and do remotes to all the different venues. 

This time, we said, no, we're gonna be everywhere. So we had six different mobile sets that allowed us to be in almost every single city to have an everywhere concept. Think about it — you want the flavor of Mexico City and Azteca Stadium, but you also want L.A. and SoFi and all the celebrities. You want the fanfare outside the stadium, that local feel, the vibrancy across North America — it’s so different from Kansas City to Miami to Vancouver. 

With all the advances in networking and storage, the cloud and AI, that’s allowed us to operate much more from the field and really been key to the flexibility and dependability that’s brought local vibes to each broadcast.

Sobel: Part of this is changes on FIFA’s part, too, where everything used to flow directly out of the IBC, the international broadcast center, right? You built up a larger facility at the broadcast center, with the tape room, the control room, so very centralized, and the technical hurdle was getting our people in and the post-production out. Now FIFA is operating in the cloud, too, so we can bring the feeds to our existing studios here in Century City, and then out to the field or to our consumer endpoints.

With this World Cup, we flipped everything on its head. We have our existing workflows in L.A., and we had to make sure we could jam the scale of the World Cup into our existing infrastructure, maybe do some cool new things, and still really make it bulletproof like Dustin said.

With all the different locations and feeds and check-in points, there's a lot of engineers with a lot of eyes on a lot of feeds. There was more than 11,500 hours of recordings. We’ve got to record them all, capture them all, tag them all, make them all accessible to the entire team, and then turn that into content at a speed and scale we never had. That translates to more than 23,000 files, 85,000 subclips, and more than 20,000 hours of live content.

That's a lot of content, to say the least, and AI is playing an increasingly important role in helping us capture all that and keep up.

Myers: Our industry is full of platitudes, but the situation really is, you've got to innovate like a startup, but be reliable, like a utility.

So can you actually walk us through the system we built together to manage and automate a lot of this tagging and search?

Myers: We had an engagement with Google a year ahead of the men’s 2018 World Cup in Russia. At the time, most of our engineers, when they were looking for a clip — not just soccer but any sport — they would just go into YouTube to find it. That would at least give them a date or time stamp, so then they could go into our internal system and narrow it from like thousands of entries down to maybe dozens or hundreds.

Sobel: It took us a few years to actually build those capabilities internally, into the IAS we have today. That’s the Intelligent Asset Service platform, launched just before the cup in Qatar, and it’s really become the hub for everything teams do at FOX Sports. Where we're at now, the combination of connectivity and bandwidth and cloud and AI tools, it’s making all this possible. Whether you’re FOX, FIFA, the BBC — you couldn't do events at this scale without a platform like IAS.

Specifically, the way the IAS works, we started on a beta beta beta version of Vision Warehouse — it was still an internal Google tool at the time, but it could “watch” the videos and really help with tagging and metadata, to accelerate that process. And then with the search APIs, it became so much easier to find clips. For now, it’s mostly keyword search, but we anticipate agentic search and conversational discovery coming online next.

There's so many ways to find media now that are just 10-times faster than before. Particularly with all the models in Gemini Enterprise Agent Platform, we’ve been able to tag and find things so much more easily and accurately. We're competing against ourselves in a sense, really just trying to make IAS and the rest of our tools as consumer-friendly as possible.

So it’s not just the expectations of the audience changing, where, how, and when they watch, but also the expectations of the crew?

Sobel: The bar for our tools isn’t set by other broadcasters — it’s set by the tech companies, like Google, whose products our people use every day. You come into work, you want the tools to work for you, not against you. You’re working 50 days on with two days off, you do it because you love being a part of these storied events and iconic moments. It’s bruising enough as it is, so we can’t let the tools beat the crew up or make things worse.

Sounds just like you said, you're innovating like a startup while staying reliable like a utility. How do you strike that balance?

Myers: There's obviously a temptation to try everything, but you can't do everything, right? It's a lot of catch-22’s. Some people expect us to have access to every tool they see. What we really do is we always pick what's going to matter to the audience or what's going to make a difference for us.

A few months before the World Cup, we sit down and talk about what’s actaully achievable that’s not going to break the team or break the broadcast. If it's a little bit of a stretch and it’ll matter to the audience, then sure. But we’re not going to risk our key assets.

Can you give us an example?

Myers: One of my favorites was with one of your partners, ElevenLabs. We did dubbing for some of the post-game interviews, but it wasn’t an interpreter, it was the players’ voices, but in English. The challenge is, even ordinarily, when players come off the pitch, we have like seven, eight minutes to turn the interviews around. And if they're in another language, it can be very hard to translate. Now, it’s near-simultaneous, thanks to what we built using the Gemini models and Speech-to-Text APIs in Gemini Enterprise.

When we talked about building it, we said we weren’t going to commit people to being here night and day for 10 days straight. But we got a beta running in four working days. A lot of that was because of the efficiencies of APIs, AI code assistants, a lot of things that are just helping speed everything up.

Myers: We’re in the storytelling business. That’s why we created Foresight with Gemini, to bring more stats and color to our MLB play-by-play teams, so they can add even more texture to the games.

Look at the history behind a game like England-Argentina, or all the match-ups between France and Spain — our commentators know these things, and our fans know these things. But three, four weeks earlier, helping them care about some of these teams they’ve never heard of, that’s where these storytelling tools can really make the difference.

So what’s next? We’ve seen some fun AI videos after the games, but how far can that go? With sports one of the last great monocultures uniting different communities, can AI further that?

Sobel: We’ll definitely keep experimenting with tools that help the crews, like the auto-transcription, the auto-highlighting that Google’s LLMs have made possible, those will keep getting better and become more and more valuable. Will we ever go to fully automated clips, I don’t think so. Like you said, sports is about the humanity of the game — the athletes, the fans, the stories, the traditions. 

Myers: It takes a human touch to stay connected to all that. We want technology that strengthens those connections.