World Cup 2026 VAR: Did AI Actually Improve Football?
- Shane Riddle
- 20 hours ago
- 19 min read
The 2026 tournament put more machine intelligence between the action and the decision than any World Cup before it. Artificial intelligence ran through the FIFA World Cup 2026 as an operational baseline rather than a novelty, with match officiating, tactical analysis, broadcasting and stadium security all leaning on it. Built out alongside official technology partner Lenovo, the tournament was widely described as the most AI-advanced sporting event in history.
I have spent a career in technology, and this is the trap I know best. We get carried away with what a system can do and forget to ask what it is actually worth to the people on the receiving end. Something can be a genuine feat of engineering and still not improve the thing it was built to serve. So now that the final whistle has gone in New Jersey, the question worth asking is not what was deployed. It is whether any of it made football, or soccer depending on where you were watching, any better. This post works through the evidence with the same lens we applied in our guide to AI in sport, separating what is measured from what is marketed, to answer one question, hype, or improvement?
Key Takeaways:
No official scorecard exists: As of early August 2026, FIFA hasn't published post-tournament refereeing statistics for 2026, There is no intervention count, no overturn count, no average review time, no accuracy figure. Any percentage you see quoted is worth checking against a primary source.
The technology answered the wrong arguments: We saw how offside margins narrowed to a toe and a hair follicle, while the decisions that actually settled matches (a penalty, a second yellow in the final) were subjective calls no sensor could resolve.
Football's own rule-makers disagreed with FIFA: On 28 July 2026, IFAB stated that the mistaken identity protocol had been misapplied in a quarter-final. FIFA publicly rejected that reading.
Fans were split, not won over: A YouGov poll of engaged British viewers during the tournament found 40% thought VAR worked well against 28% who thought it worked badly.
The clearest winners were vendors, not the sport: Lenovo's own communications lead called the tournament "the most valuable real estate in the world", and a mapping investigation identified at least 21 companies that built or expanded surveillance infrastructure around it.
The trickle-down is real but unpriced: Football Video Support, FIFA's low-resource review system, is available now in Italy's Serie C, Spain's Liga F, the Canadian Premier League, the Australian Championship and, from August 2026, the USL. Nobody involved has published what it costs.
Table of Contents
The acronyms, in plain English
Football officiating has collected a lot of initials over the past decade. Here is what each one means, so nothing below needs decoding.
Acronym | Stands for | What it actually is |
VAR | Video Assistant Referee | Officials watching replays away from the pitch who can advise the referee when they spot a clear and obvious error. Covers four things only: goals, penalties, direct red cards, and mistaken identity. |
SAOT | Semi-Automated Offside Technology | The AI layer that tracks players and the ball to reconstruct exactly where everyone was at the moment of a pass. It produces the evidence; a person still makes the call. |
FVS | Football Video Support | FIFA's stripped-back review system for competitions that cannot afford full VAR. The coach challenges a decision, twice a match maximum. |
IFAB | International Football Association Board | Writes the Laws of the Game. No technology can be used in official competition without its approval. |
FIFA | Fédération Internationale de Football Association | Football's world governing body. Organises the World Cup and pays for the technology used at it. |
AR | Augmented reality | Graphics laid over a live video picture, such as an offside line drawn onto the broadcast feed. |
USL | United Soccer League | A professional league structure in the United States, trialling FVS from August 2026. |
What AI actually did at the tournament
Before judging the impact of the World Cup 2026 VAR, it helps to be precise about what was running. All of the following were available now at the 2026 tournament, confirmed by FIFA or its technology partner.
Advanced semi-automated offside technology (SAOT). FIFA confirmed 16 optical tracking cameras in each of the 16 stadiums, generating over 150 million tracking data points per match. The significant change from Qatar 2022, clear offside alerts went directly to the on-pitch officials rather than being routed through the video assistant first.
Personalised 3D player avatars. Every player was body-scanned before the tournament, with Lenovo reporting 1,248 scans completed. The avatars replaced the generic silhouettes used in earlier systems and drove the animated offside reconstructions shown on stadium screens and in broadcast.
Connected ball technology. The adidas Trionda carries a Kinexon inertial sensor transmitting 500 times per second, timestamping the moment of contact. FIFA described the sensor's output to viewers as a "heartbeat graphic".
Referee View. An AI-stabilised body camera worn by the referee. Both Lenovo and FIFA present it as a fan and broadcast feature, and no source we found describes it being used to make officiating decisions. FIFA's refereeing chief Pierluigi Collina framed its primary value differently again, as a coaching tool: seeing what the referee saw is "important in the debriefing".
FIFA AI Pro. A match analysis platform made available to all 48 squads, analysing over 2,000 performance metrics. Notably, Axios reported that FIFA prohibited its use during live play, restricting it to pre- and post-match work.
Security and operations AI. Crowd pattern analysis, digital twins of venues, counter-drone systems and an intelligent command centre spanning FIFA's operational functions.
What was not present was any autonomous AI referee. Every automated alert still passed through a human before it became a decision. That distinction is important for everything that follows in our review.
Verdict 1: Did decisions get more accurate?
This is the question the technology exists to answer, and it is the one where we found the evidence to be the thinnest.
The missing scorecard
We waited for several weeks before creating this post but todate, FIFA hasn't published any post-tournament refereeing statistics for 2026. Collina's mid-tournament statement after the 72 group-stage matches covered time-wasting and discipline (4 goal-kick violations, 11 throw-in violations, 10 red cards) and contained no VAR metrics at all. A second statement after 96 matches was in defence of integrity with no statistics whatsoever. FIFA's post-tournament release celebrated attendance, viewership and goals, and said nothing about officiating.
That absence is worth sitting with for a moment. After Russia 2018, FIFA published a headline VAR accuracy figure but after 2026, it didn't. Percentages currently found circulating online, including a 99.3% accuracy claim attached to this tournament. This appears to trace back to that 2018 release rather than to any 2026 source.
The most rigorous independent measurement came from Brennan Klein at Northeastern University's network science institute, who normalised VAR decisions per 100 fouls through the round of 16. The analysis was published by Northeastern and reported by The National. Klein's own caveat is the important part, the numbers show where referees missed fouls the VAR later flagged, not proof of bias.
The rule-makers and FIFA fell out in public
The clearest evidence that something went wrong isn't a fan complaint, it was a disagreement between the two bodies that run the sport.
In the quarter-final between Argentina and Switzerland on 11 July, VAR reviewed a foul on Breel Embolo, decided he had gone down before contact, rescinded the card given against Argentina's Leandro Paredes and instead showed Embolo a second yellow for simulation under the new mistaken identity protocol. Switzerland played on with ten men and lost in extra time. Switzerland manager Murat Yakin said his side had been punished "because of a rule that, in my opinion, is completely unacceptable".
On 28 July, IFAB stated that the protocol had been misapplied in that a card can be reviewed to identify who committed the offence, but the offence itself cannot be re-decided. FIFA rejected that characterisation the same day, saying its interpretation had been applied consistently throughout the tournament and that it did "not consider this to have been a referee or VAR error". The VAR protocol is now up for revision.
Did the technology get very good at answering questions nobody was asking?
Three decisions from the tournament illustrate the pattern.
Colombia's Davinson Sánchez had a stoppage-time header against Portugal ruled out for offside by a toe. His coach Néstor Lorenzo's response, reported by Newsweek, was "Next time, I'll tell him to get a pedicure."
Croatia went out of the tournament in part because the ball sensor detected contact in the build-up to a disallowed equaliser against Portugal. FIFA explained that the Trionda's sensors "are capable of determining any slight contact". The contact was reported to have come off a player's hair. Croatia manager Zlatko Dalić: "All these decisions take the joy out of football."
Germany went out to Paraguay after an extra-time goal was disallowed for an off-the-ball block. Julian Nagelsmann called it "not just a scandal, it's an absolute scandal". Jürgen Klopp, watching on German television, made the sharper point: "If the goal is illegal, then Arsenal won't be English champion. They've scored 60% of their goals that way."
None of those three was a technology failure. The tracking worked. What each exposed is that measurement is not judgement, and that the sport had not agreed on where to draw the line before it built a system precise enough to make the line matter. Collina's own doctrine, that "a foul is a foul" with no defined distance or time limit on how far back in a possession phase the VAR may reach, is a policy choice, not a technical capability.
Portugal manager Roberto Martínez put the case for the defence: "The balls now have a chip, and it's very clear that's why the VAR intervened. It's not a subjective opinion." He is right that the data was not opinion. The argument was always about whether that particular fact should decide a World Cup match.
And the tournament's two most consequential decisions were not technology calls at all. The semi-final penalty that broke France, and Enzo Fernández's second yellow card three minutes into stoppage time in the final, were both judgement calls by a human official on an incident no sensor could adjudicate.
The most useful summary of all this came from FIFA's own refereeing chief. Speaking to TechRadar on the eve of the final, Collina said:
"The misconception is we have the technology, and we can provide an answer to everything." — Pierluigi Collina, FIFA Chief Refereeing Officer
My Verdict: Not proven. The system almost certainly caught errors that would once have stood. It also produced a tournament remembered for officiating controversy, and FIFA declined to publish the numbers so far that would settle the argument either way.
Verdict 2: Did watching football get better?
The clock
Match length is the fan complaint the technology was supposed to address, and the data here is frustratingly partial. The figures in circulation (an average match length of 96 minutes 8 seconds in 2026 against 102 minutes 43 seconds at Qatar 2022, with ball-in-play time of 57 minutes 22 seconds) were reported one week into the tournament, not after it, and no data provider is named as their source. No full-tournament recalculation has been published.
Two caveats are important if you plan to quote them. First, the reduction is credited to Collina's crackdown on time-wasting (five-second limits on goal kicks and throw-ins, ten seconds for substituted players to leave the field), not to faster VAR. Second, those averages exclude the mandatory hydration breaks introduced in 2026, so the time a viewer actually sat through did not fall by anything like six minutes. One reporter covering the fan experience described the tournament as having become "a game of four quarters".
What fans said
The only independent polling we could find is a YouGov survey of 2,052 British adults conducted 6 to 7 July 2026. Among the 816 respondents following the tournament a great deal or a fair amount, 40% said VAR worked well (4% very well, 36% fairly well) and 28% said it worked badly. A meaningful margin, but a long way from vindication, and the poll asked nothing at all about the AI broadcast features.
That gap is interesting and somewhat telling in that the only quantitative data on how fans received Referee View, the 3D avatars and the AR graphics comes from a survey Lenovo commissioned and conducted itself. Lenovo surveyed 2,514 adults across five countries, fieldwork 24 to 27 June, no independent polling house named. It reports that 91% of fans found the broadcast technology made the viewing experience more dynamic and 87% said technology improved their viewing. We found that those numbers were then republished across a dozen outlets, which gives one press release the appearance of a consensus.
The most interesting praise the technology received
The best independent verdict came not from a fan survey but from an industry analyst, and it is a compliment that cuts both ways. Reviewing the tournament's immersive technology, UC Today quoted Argus HD founder Tim Kay:
"The tech finally got boring, and I mean that in the absolute best way possible." — Tim Kay, Argus HD
I know from my own work of when tech works the way its intended, that this is a genuine praise for engineering that worked. However, it's also the opposite of what the tournament was sold as. And it lands awkwardly next to the claims that Lenovo's post-tournament release cites "over 20 billion impressions" in connection with the Referee View footage of the opening goal, while trade publication Sports Video Group reported that the ref cam clip from Mexico against England drew 90,000 views on FIFA's YouTube channel. Both figures are accurately reported, they can't both describe the same thing.
Collina's defence of the wait before a goal is confirmed deserves a hearing, because it is the strongest version of the pro-technology case:
"The player immediately reacts and celebrates the goal. And then there is a moment of suspense. And at the end, when the goal is confirmed, there is a second celebration. So basically, the emotion is doubled; it's not reduced." — Pierluigi Collina, TechRadar
Not everyone experienced it that way. Felipe Cardenas of The Athletic told NPR that VAR "is completely interrupting what the game state is", adding "As football fans, you should live with human error at times."
My Verdict: Mixed, and largely unmeasured. Matches ran shorter, for reasons that had little to do with AI. The AI broadcast layer worked well enough to go unnoticed, and nobody independent has asked fans what they made of it.
The aspect that AI has significantly altered is the creation of fakes.
In my research, I found there is one area where artificial intelligence had a large, documented and entirely unhelpful effect on the World Cup, and it had nothing to do with FIFA's technology stack and everything to do with 3rd parties monetisation of the event. The European Digital Media Observatory documented a wave of AI-generated images of hypersexualised female supporters who do not exist, verified as generative AI output by fact-checking organisations across Belgium, Germany and Italy, and monetised through platform creator programs.
In the days after the final, a consortium of European public broadcasters including VerificaRTVE, SRF, France 24, VRT and DW ran a joint fact-check that identified AI-generated images of Lamine Yamal with the referee, staged to imply match-fixing, alongside several other fabricated images and a fabricated managerial quote about the officials.
Futurism documented AI-generated clips circulating on X with over seven million views, and a fabricated image of two fans celebrating a Sweden goal that was never scored.
This is worth noting in any assessment of the tournament's AI credentials. The technology FIFA controlled behaved as designed. The technology nobody controlled spent five weeks manufacturing things that never happened, and European public broadcasters spent the tournament debunking them.
Verdict 3: Who actually benefited?
Follow the money and the picture gets clearer than the officiating data ever does.
Lenovo has never disclosed what the FIFA partnership cost. What its executives have said on the record is more useful. Speaking to SportsPro in June, Lenovo's head of communications Jeff Shafer described the arrangement in terms that leave little ambiguity:
"This is the most valuable real estate in the world. For seven minutes in every game you're going to see Lenovo and Motorola [branding]." — Jeff Shafer, Lenovo
Shafer told Forbes the same thing in different words: "This is legacy work for the company. This is the kind of thing that changes the brand of the company for a long time." None of this is improper. It is simply a clear statement of what the tournament was for, from the party that paid for it.
The security layer raises harder questions. A mapping investigation by Ranking Digital Rights, No Fair Play, identified at least 21 companies that built or enhanced surveillance infrastructure for the tournament, most of them winning new contracts specifically for it. US host-city security grants through FEMA totalled $625 million. Facial recognition operated at stadiums in Atlanta and Boston with automated licence plate readers that were deployed in the hundreds.
The report's most useful finding for anyone thinking about legacy is about what happens next. Of 16 host cities, only 13 published human rights action plans, only three addressed privacy explicitly, and only Vancouver committed to an explicit decommissioning plan for its cameras. Atlanta removed the privacy section from its final plan. Kansas City postponed, rather than cancelled, facial recognition on its public buses.
Meanwhile Gianni Infantino said FIFA expects to "top the USD 15 billion mark" in revenue for the 2023 to 2026 cycle. Within a fortnight of the final it proposed selling up to 20% of a new commercial subsidiary to outside investors, a plan UEFA called "irresponsible and indefensible" and which FIFA scrapped on 31 July after a confederation revolt.
My Verdict: The vendors, clearly. The sport got a technology showcase. The host cities got infrastructure that mostly stays switched on.
Verdict 4: What it means for your local club
This is where the picture improves, and it is the part of the story that got least attention.
The system worth watching is not SAOT. It is Football Video Support (FVS), FIFA's challenge-based review system built for competitions with fewer cameras and less money. Only the head coach can request a review, by twirling a finger and handing a card to the fourth official. Each team gets two requests per match. Four categories are reviewable: goal, penalty, direct red card, mistaken identity. There is no team of video officials monitoring every incident.
Collina has been careful about the framing: "FVS is a tool to support referees in competitions with fewer resources and cameras. It should not be seen as VAR or as a modified version of it."
FVS is available now in a growing list of competitions:
Italy's Serie C, with over 200 matches played under FVS as of October 2025
Spain's Liga F and Primera Federación
Brazil's Copa Paulista and Copa do Brasil Feminina
The Canadian Premier League, from April 2026
The Australian Championship, across its finals series in November 2025
The USL, which confirmed a trial across the final seven matches of the 2026 USL Cup, starting 12 August 2026
USL President of Competition and Administration Brett Luy put the case for it directly: implementing FVS "gives us an opportunity to help evaluate a technology that could make high-level officiating more accessible."
My research has found there is one catch, and it is a significant one. No organisation involved has published what FVS costs. Not FIFA, not Hawk-Eye, not Dartfish, not Spiideo, not a single league using it. Every description uses comparative language ("cost-effective", "affordable", "without the hefty costs of full VAR") with no figure behind it. For a system whose entire pitch is affordability to resource-poor competitions, that is a strange gap. The only fully costed comparator we could find is dated with the Scottish Premiership clubs being quoted close to £120,000 per club per season for full VAR back in 2022.
There is also a layer beneath the technology that rarely appears in the sponsor communications. Reporting by Tatiana Dias for Tech Policy Press documented the human annotators in the Philippines, Cambodia, Laos and Brazil who manually code match footage that feeds these systems, including one annotator in Rio de Janeiro paid roughly €60 per match and financially penalised for late submission. Her structural point is an important one for community sport in that most clubs cannot collect data independently, which is exactly what makes the data-vendor layer profitable. As we found looking at what the sports technology boom actually delivers to community clubs, the gap is rarely about hardware. It is about who can afford the ongoing subscription.
My Verdict: genuine progress, poorly explained. If you run or play in a semi-professional competition, FVS is the development to watch. Ask for a price before you get excited though.
What this looks like from Australia
Australia has a claim in this story that rarely gets made, it went first.
On 7 April 2017, Melbourne City played Adelaide United at AAMI Park in what Football Australia describes as the first use of VAR in a top-tier national competition anywhere in the world. IFAB secretary Lukas Brud said at the time that the lawmakers were "very pleased with the very professional preparatory work done in Australia". The first decision actually overturned came the following night, when a waved-away handball in Wellington Phoenix against Sydney FC became a penalty.
Nine years on, the country that went first had a World Cup goal given against it by the technology it helped prove. In Seattle on 20 June, Alex Freeman's 43rd-minute goal for the United States was flagged offside on the field and then awarded after a VAR check. SBS put it plainly, one own goal, "the other an offside decision overturned by technology". Australia lost 2-0, drew with Paraguay to reach the knockouts for only the third time in their history, and went out to Egypt on penalties.
Australians were on the other side of the decisions too. Football Australia had four officials at the tournament with Alireza Faghani as a referee, George Lakrindis and Andrew Lindsay as assistants, and Shaun Evans as a video match official. Faghani ended up in the middle of the tournament's loudest refereeing row, England's 3-2 win over Mexico, after which Thomas Tuchel said VAR had overturned a decision that was "absolutely not" a clear and obvious mistake.
The numbers FIFA did not publish, and Football Australia did
Here is the comparison worth noting, six weeks after its season finished, Football Australia published an officiating report: 91% overall decision accuracy in the A-League Men, 97% in the A-League Women, 38 VAR reviews across the season, roughly one every four matches, and VAR errors halved from 18 to nine.
FIFA, six weeks after a tournament it called the most AI-advanced in history, has published none of the equivalent figures. One of those bodies has a tiny fraction of the other's resources. Transparency here is not a question of capability.
Where the technology sits below the A-League
Australia is further along at the cheap end than most countries as well. The Australian Championship, the new national second division, was among the first competitions FIFA selected to trial Football Video Support, using it across the finals series in November 2025: two coach challenges per team, a pitch side monitor, no remote hub.
Jon Moss, Football Australia's head of referees, made the case in a sentence that applies well beyond Australia. FVS "has been designed precisely for a competition like ours, where broadcast coverage is strong but not extensive enough for a full VAR set-up".
Below that, there is nothing. No review technology in the National Premier Leagues or the Australia Cup.
What community football has instead is cameras bought for other reasons. Football NSW rolled out Veo units across its state teams, NPL clubs and youth leagues in 2025, for coaching and streaming. Football Australia's grassroots body camera pilot put 45 cameras into Queensland, New South Wales and Victoria, aimed at referee retention and behaviour towards officials rather than at reviewing decisions. The cameras are already on the sidelines. Nobody has connected them to the whistle.
My Verdict: Australia proved the case for VAR nine years before this tournament, and is now trialing the cheap version of it. What nobody has done, here or anywhere else, is say what either one costs the clubs below the top tier.
The scorecard
Did decisions get more accurate? Not proven. FIFA published no 2026 accuracy figures, intervention counts or review times. IFAB and FIFA publicly disagreed over one quarter-final decision.
Did anyone publish the numbers? FIFA didn't, Football Australia published a full officiating report for its own season, decision accuracy included, six weeks after it ended.
Did matches get shorter? Probably, but credited to time-wasting rules rather than AI, and the only figures available cover the first week and exclude hydration breaks.
Did fans approve? Split. 40% positive against 28% negative among engaged British viewers. No independent polling exists on the AI broadcast features at all.
Did the broadcast improve? Modestly, and mostly invisibly. The strongest independent praise was that the technology "finally got boring".
Did teams benefit? Unknown. FIFA AI Pro went to all 48 squads, but no coach, analyst or federation is on record evaluating it.
Did anyone clearly gain? Yes: technology vendors, and the security firms whose infrastructure stays in host cities after the tournament.
Will it reach grassroots? Partly. Football Video Support is available now in several lower-tier leagues. No supplier has published a price.
Final Thoughts
The engineering at this World Cup was real. A one-second body scan producing an avatar accurate enough to adjudicate a shoulder, a ball reporting its own position 500 times a second, a chest-mounted camera stabilised well enough to broadcast, none of that is a concept video. It worked, at scale, across 16 stadiums and 104 matches.
What the tournament did not produce is evidence that any of it made football better. FIFA published the attendance figures, the viewership figures and the revenue figures, and stayed quiet on the one set of numbers that would have answered the question. The rule-makers and the organisers ended up arguing in public about a decision the technology helped produce. And the calls that decided the biggest matches were, as they have always been, one person's judgement about what they just saw.
None of this is an argument against technology in sport. It is an argument for asking better questions of it. The most valuable thing to come out of 2026 for the vast majority of people who play and watch football is probably Football Video Support. Two challenges a match, four reviewable categories, no permanent camera rig, arriving now in third tiers and women's leagues where a wrong call has never had a remedy. Australia's own second division was among the first competitions FIFA picked to try it. That is a smaller idea than a digital twin of a stadium. It is also the one most likely to reach your club.
The referee is not being replaced. Nor, on this evidence, have they been meaningfully upgraded. What changed in 2026 is the volume of information around the decision, and the number of people with a commercial interest in you finding that impressive.
Frequently Asked Questions (FAQ)
Q: Did FIFA publish accuracy statistics for VAR at the 2026 World Cup?
A: No: as of early August 2026 no post-tournament refereeing statistics had been released, covering intervention counts, overturns, review times or accuracy. Accuracy percentages circulating online in connection with 2026 appear to originate from FIFA's 2018 World Cup release. Check the source before repeating them.
Q: Was the offside technology at the 2026 World Cup fully automated?
A: No: it remained semi-automated. The significant change from Qatar 2022 was that clear-cut offside alerts went directly to the on-pitch officials rather than through the video assistant first. A human still made every final call, and no fully automated system has been approved for competitive use.
Q: Did AI make World Cup matches shorter?
A: Not directly. The reduction in average match length reported early in the tournament was attributed to new time-wasting rules, including five-second limits on goal kicks and throw-ins. Those figures also exclude the hydration breaks introduced in 2026, so the elapsed time a viewer sat through fell by considerably less.
Q: What is Football Video Support and how is it different from VAR?
A: FVS is a challenge-based review system for competitions with fewer cameras and smaller budgets. The head coach requests a review (two per team per match) and only four categories are reviewable, goal, penalty, direct red card and mistaken identity. Unlike VAR, no video officials monitor every incident. It is available now in Italy's Serie C, Spain's Liga F, the Canadian Premier League, the Australian Championship and, from 12 August 2026, the USL Cup.
Q: How much does officiating technology cost a lower-league club?
A: No supplier or league has published a price for Football Video Support, despite affordability being its central claim. For comparison, Scottish Premiership clubs were quoted close to £120,000 per club per season for full VAR in 2022. If your competition is considering FVS, ask for a costed quote including ongoing fees rather than relying on the word "affordable".
Q: Was the referee body camera used to make decisions?
A: No: FIFA and Lenovo both present Referee View as a fan and broadcast feature, and no source describes it being used to make officiating decisions. FIFA's refereeing chief Pierluigi Collina described its main value as post-match debriefing, letting assessors see what the referee saw when a call was made.
Q: Did AI create problems at the World Cup that FIFA did not cause?
A: Yes: European fact-checking organisations documented significant volumes of AI-generated fake content around the tournament, including fabricated images of players staged to imply match-fixing and AI-generated images of supporters who do not exist. Public broadcasters across Europe ran joint fact-checks through the final.
Category: Technology & Sport
Related Posts:
