GPT-5 and Large Language Models:
How AI is Rewriting Professional Football Tactics

In May 2026, something fundamentally changed in the quiet offices of European sporting directors. More silently than a missed penalty but with infinitely more consequences, a new era has opened: one where GPT-5 and large multimodal language models no longer just analyze data, but are truly beginning to think about football. Formulating strategies. Questioning the decisions of coaches who have spent twenty years on the pitch. This is no longer science fiction. It is the reality for clubs like Chelsea, Liverpool, and Bayer Leverkusen, pioneers of a revolution disrupting every link in the sports chain.

For decades, football resisted data science. Too human, too unpredictable, too dependent on instinct. Zlatan Ibrahimović scored from impossible angles; Zinedine Zidane read a press before it even formed. Can this really be captured in a statistical model? The answer, for a long time, was no. But GPT-5 is not an ordinary statistical model. It is a multimodal reasoning system capable of simultaneously ingesting video feeds, GPS data, medical reports, match histories, and psychological profiles — extracting patterns that the human eye, even the most expert, is unable to perceive.

Professional football is entering a new era. And those who do not adapt risk simply disappearing from the top of the table.

From Video to Multimodal Model

For a long time, video analysis in football was a matter of human eyes. Analysts spent hours watching game sequences, annotating clips, laboriously building PowerPoint presentations for the coach. It was valuable, but it was slow, subjective and above all — limited to what a human brain could reasonably process.

GPT-5, in its multimodal version deployed in early 2026, has changed the game spectacularly. The model can now analyze hours of footage in real time, identify pressing patterns, calculate dynamic spaces between lines, and generate comprehensive tactical reports in natural language — in seconds. What used to take a team of analysts three days now takes ten minutes for a system like StatsBomb Intelligence, which integrates GPT-5 in the backend.

📊 Computer vision systems coupled with LLMs turn every match into a flood of actionable data — GPS positions, heatmaps, trajectory predictions. © Unsplash

The revolution does not stop there. Next-generation multimodal models are capable of cross-referencing video feeds with real-time physiological data. The midfielder's GPS pedometer, heart rate, estimated muscle fatigue by predictive algorithm — all this is integrated to generate substitution recommendations or positional adjustments at half-time. We are now talking about "augmented coaching" : the coach retains the final decision, but his intuition is enriched by an analytical layer unprecedented in the history of sport.

Tools like Hudl Sportscode, Wyscout, or even proprietary solutions developed internally by big clubs now integrate APIs connected to LLMs. These pipelines can transform a raw 90-minute video into a structured tactical synthesis, with opposition recommendations, alerts on opponent trends, and even simulations of alternative scenarios ("what happens if we switch to a defensive 4-3-3 in the 60th minute?"). The answer arrives in real time. The dugout of the future looks more like a military command post than a dressing room.

Algorithmic Scouting Revolutionizes Transfers

While tactical analysis has fascinated the general public, it is in the transfer market that AI produces its most economically significant — and radical — effects. The professional football transfer market is worth several billion euros a year. Every recruitment error not only costs a fortune but can destabilize a dressing room for entire seasons. The emergence of scouting tools powered by LLMs is profoundly transforming this process.

🔍 Sporting directors increasingly rely on AI dashboards to identify talent before their competitors. © Unsplash

Imagine a sporting director asking a GPT-5 based system: "Find me a right-back under 24 capable of playing in a high 4-2-3-1, with a pressing rate of over 8 intense sprints per match, an interception ratio in the final third of over 2.3, and a transfer cost of under 20 million euros." In under a minute, the system returns profiles of players from 47 different leagues, with contextualized analysis, career trajectory comparisons, and even estimates of cultural compatibility with the target dressing room.

Platforms like Transfermarkt Intelligence or Opta AI Scout have integrated natural language processing layers to enable conversational searches. The traditional scout — who spent weekends in regional third-division stadiums — is not disappearing, but their role is radically changing. They become a human validator of machine-generated hypotheses, a cultural and emotional filter that the algorithm cannot yet replicate. Dressing room dynamics, a captain's charisma, psychological resilience to adversity — these dimensions remain human territory.

Chelsea, Liverpool, Bayer Leverkusen: The Pioneers

Three clubs symbolize this quiet but profound revolution better than any other. Their approaches differ, their results speak for themselves, and their example inspires — or terrifies — the rest of European football.

🏟️ The backstage areas of major European stadiums now house data labs as important as the weight rooms. © Unsplash

Chelsea FC: The Premier League Data Lab

Since the takeover by the BlueCo consortium in 2022, Chelsea has invested heavily in a data infrastructure unmatched in the Premier League. The London club partnered with an American startup to develop a system called "BlueIntel", which integrates a large language model fine-tuned on ten years of proprietary data: matches filmed from 12 angles, biometric data of the entire professional squad, anonymized medical reports, and tactical exchanges transcribed during training sessions. The result? A system capable of automatically generating "opposition tactical memos" before each match, summarizing opposing teams' playing patterns with a precision that stuns human analysts. Internally, coaches admit that BlueIntel identifies opponent pressing trends they would only have noticed after three or four games of observation.

Liverpool FC: When Expected Goals Evolve into Expected Intelligence

Liverpool has always been an avant-garde club in terms of data, since the Ian Graham era and the famous research department that reportedly influenced the signing of Mohamed Salah. Today, under the direction of its new head of innovation, the club has taken another step by deploying a conversational AI agent accessible directly from the coach's iPad during matches. This system, powered by an adapted version of GPT-5, can answer natural language questions in real time: "Is our pressing effective in this first quarter-hour?" or "Which opposing player is creating the most imbalance in our defensive block?". The answers, contextualized and documented by the current match data, arrive in 3 seconds. Anfield enters the era of conversational coaching.

Bayer Leverkusen: AI at the Service of Total Football

The Rhineland club is perhaps the most striking example of this fusion between playing philosophy and artificial intelligence. The historic 2023-2024 season, concluded unbeaten in the Bundesliga under Xabi Alonso, had already revealed a tactical organization of rare coherence. What the public didn't know then was that the Bayer staff was already using sophisticated predictive analysis tools. In 2026, the team took another step: its data science department, composed of seven engineers, developed a predictive fatigue analysis system coupled with an LLM. It continuously analyzes player workload and generates squad management recommendations over a three-match window — optimizing rotations even before signs of physical fatigue become visible. Football becomes a high-precision human resources management science.

The Ethical Limits of Augmented Coaching

Fascinating as it is, this revolution is not without fundamental questions — ethical, human, and even philosophical — about the very nature of sport. For while AI can optimize a squad, predict an injury, or identify a hidden talent in a Romanian third division, it relies on foundations whose fragilities deserve rigorous examination.

The first dividing line is player privacy. Continuous biometric monitoring systems — heart rate, sleep quality, estimated hormonal stress levels — generate extremely sensitive volumes of personal data. Who accesses this data? Can it influence a transfer or contract renewal decision? In 2025, the international players' union FIFPRO published an alarming report revealing that in 23% of the clubs studied, individual biometric data was accessible to sporting decision-makers without the explicit consent of the players concerned. The French CNIL and its European equivalents are beginning to tackle the subject.

The second question concerns algorithmic determinism. If an AI system predicts a player has a 78% chance of a knee injury in the next six weeks, should they play? Can they access this information themselves? And if the prediction is wrong — as any probabilistic model inevitably will be — who bears the responsibility for the decision made on that basis? These questions are not abstract. Several cases have already emerged in the Italian Serie A and the Spanish La Liga, where players challenged their benching by citing that an algorithmic decision had supplanted the coach's sporting judgment.

Finally, there is the question of competitive fairness. If clubs with the biggest budgets can afford the best AI systems — and the data scientists to run them — the gap with less well-endowed teams risks widening irreversibly. Football, already profoundly unequal financially, could see this asymmetry transform into a true technological chasm. The Premier League is actively discussing data sharing rules and mutualized access to certain platforms for bottom-half clubs. UEFA and FIFA, in turn, are working on regulatory frameworks for the use of AI in professional sport — a titanic project of which we currently only see the first milestones.

Conclusion: The Future is Now

GPT-5 and large multimodal language models are not a threat to football. They are its next natural evolution. Just as television transformed stands into a global audience, just as 2010s data analytics revolutionized recruitment, 2026 AI is redrawing the contours of what it means to prepare for a match, manage a squad, and build a competitive team.

But this revolution is neither fatal nor blind. It calls for intelligent governance, solid protections for the game's actors — starting with the players themselves — and a collective reflection on what we want sport to remain: a space for unpredictable human genius, enhanced by technology, and not reduced to an optimization equation.

The clubs that win in this new era will not necessarily be those with the most powerful algorithm. They will be those who have intelligently integrated AI into a strong sporting culture, serving clear human values. Chelsea, Liverpool, and Leverkusen have understood this. It is up to others to follow — or be overtaken by this wave that nothing, now, seems able to stop.