The 2026 World Cup is already under way, and the commentary is live again in every market. So before this tournament writes its own bloopers, here is a reminder of what the booths got wrong last time, and which of those mistakes never had to happen.
Commentary is hard, and the people who do it well are extraordinary. This is not an argument that humans are bad at it. It is an argument that some kinds of mistakes are avoidable by design, and those are the ones worth talking about two weeks before another World Cup begins.
Here are five from 2022, ordered from the most fixable to the least.
1. The Booth Told Viewers the Opposite of What Was Happening
During Spain against Japan, the English-language broadcast told its audience that Spain would be eliminated if they lost and Germany won, and that Spain needed an equaliser to go through. Spain did lose. Germany did win. And both Spain and Japan advanced anyway, with Germany going out on goal difference.
The math here was knowable in advance. Every qualification scenario for every scoreline can be computed the moment a group's permutations lock in. Liat.ai loads those projections before kickoff, so when a goal goes in, the standings update is already correct. A commentary engine built this way cannot tell millions of people the reverse of the truth, because it is reading a table, not a memory.
2. The Bracket on Screen Was Wrong
After the groups finished, the live broadcast showed a last-sixteen bracket with the wrong fixtures, pairing teams that were not actually due to meet. The correct matchups were fixed and verifiable the instant the final group results were confirmed. This is the simplest category of error there is, a data-entry mistake on live television, and it is the easiest to remove entirely.
3. Player Names Were Mispronounced
Before the Germany match, well-known names including Neuer, Müller and Gnabry were stumbled over on air. Pronunciation is not a small thing. It is the most basic respect a broadcast owes a player and an audience. Liat.ai builds a native commentary pair for each language, so a German name is pronounced the German way in the German feed and correctly in every other feed too. The pronunciation is built in, not improvised.
4. Commentary Bias Is Real, and It Is Measurable
In 2020, the Professional Footballers' Association worked with the research firm RunRepeat to analyse 2,074 statements made by English-language commentators across 80 matches. The findings were stark. Commentators were 6.59 times more likely to talk about power when describing a darker-skinned player, and 3.38 times more likely to talk about pace. More than 60 percent of praise about work rate, and a similar share of praise about intelligence, went to lighter-skinned players.
This is the kind of bias no individual sets out to produce. It emerges from habit, from language, from who is watching whom. A generation of broadcasters has been asked to correct it through training, and that work matters. But Liat.ai can do something a human booth cannot. Every word our commentary uses is logged. A broadcaster can run a report and see exactly how often "powerful" or "intelligent" was applied to each player, and hold the output to a consistent standard across every name. The bias does not just get discouraged. It becomes audit-able.
5. Refereeing Decisions, Where We Are Honest About the Limits
The 2022 tournament had its share of refereeing controversy, from the millimetric video review that allowed Japan's winner against Spain to the record eighteen yellow cards in Argentina against the Netherlands. We will not pretend AI commentary fixes this. Refereeing is a judgment call, and judgment is human territory.
What Liat.ai can do during any video review is useful all the same. It can instantly surface the relevant law text, the comparable past rulings, and the live effect of the decision on the group standings, in the language you are watching in, while the review is still happening. It does not replace the referee's call. It just makes sure you understand it.
The Pattern
Four of these five errors share a root. They are failures of recall, of data entry, of consistency, and of preparation under time pressure. Those are precisely the things a real-time engine does not struggle with. The fifth, refereeing, stays human, and we think it should.
Great commentary will always need human feeling. But the avoidable mistakes do not have to be part of the package. That is the part we built Liat.ai to take off the table.
Sources: World Soccer Talk and Sports Media Watch (FOX Sports Qatar 2022 coverage review); RunRepeat and PFA "Racial Bias in Football Commentary" study, 2020, as reported by CNN and Sky Sports; ESPN match records for the Japan goal-line review and the Argentina v Netherlands card record.
Hear how Liat.ai calls a match on the demo page, read more about the system on the technology page, or get in touch about coverage.