By
Stuart Kerr, Technology Correspondent, LiveAIWire
At the 2024 Paris Olympics, AI systems monitored gymnastics
routines from multiple camera angles, providing judges with biomechanical
breakdowns of each movement within seconds of completion. The technology was
presented as a judging aid — a tool to reduce inconsistency and human
error in a sport long troubled by scoring controversies. Critics called it
the beginning of something harder to reverse: the algorithmic audit of
athletic achievement.
Sport has always been in tension with the technologies available
to measure, enhance, and adjudicate performance. AI is sharpening that
tension in every direction simultaneously
— changing how athletes train,
how performances are judged, how doping is detected, and how governing bodies
define what counts as fair competition in an era when the boundary between
human and technological enhancement is less clear than it has ever
been.
Training on Data: AI as Performance Engineer
Elite athletes have access to performance data at granularity that
would have been inconceivable a generation ago. Wearable sensors capture
biomechanical data at hundreds of data points per second. Video systems track
movement patterns across every training session. AI models trained on this
data can identify micro-inefficiencies in running gait, swimming stroke, or
weightlifting technique that are invisible to the human coaching eye and
which, when corrected, compound into meaningful performance gains over months
of training.
The competitive advantage of AI-optimised training is real, but it
is distributed unequally. National federations with large budgets —
the United States, China, Germany, Great Britain —
can invest in comprehensive AI training infrastructure. Athletes from
smaller federations cannot. The result is an emerging technological
performance gap that sits alongside the economic one that has always
disadvantaged smaller nations, compounding rather than correcting existing
inequalities in Olympic competition.
What this means for you as a sports fan: the performances you see
at elite level are increasingly the product of human-AI collaboration. The
question of how much credit belongs to the athlete versus the training
technology is not yet a mainstream debate in sports culture, but it is one
that governing bodies will eventually be forced to address.
AI Judging and the Objectivity Claim
The appeal of AI in subjective judging —
gymnastics, figure skating, diving
— is the promise of
consistency. Human judges exhibit in-group bias, order effects, and fatigue;
they are influenced by crowd noise and national reputation. An AI system that
scores each routine against the same biomechanical criteria in the same way,
every time, appears to solve problems that have generated Olympic controversy
for decades.
The difficulty is that the criteria themselves are contested. What
counts as a “clean” landing or a “complete” rotation is
not a neutral technical fact but a judgement encoded in the training data. If
the training data reflects the preferences of historically dominant national
schools of gymnastics, the model will inherit those biases while appearing to
apply objective standards. Research from sports science journals has noted
that AI scoring systems can exhibit systematic variation across athlete body
types — rewarding body proportions common in
historical champions and penalising those less represented in training
data.
The World Anti-Doping Agency has separately been investing in AI
tools to identify doping patterns that elude conventional testing —
analysing longitudinal biological passport data, performance
trajectories, and physiological markers to flag athletes for targeted
testing. Research published by WADA
has described machine learning approaches that detected anomalies in
biological passport data consistent with microdosing EPO — a
practice that conventional sampling misses because it uses doses too small
and timed too precisely to appear in standard tests.
The Equipment Arms Race
AI is also transforming the design of athletic equipment in ways
that challenge sport’s traditional definitions of fair competition. Nike’s
Vaporfly and Alphafly running shoes, whose carbon-fibre plate designs were
optimised using AI-assisted aerodynamic and biomechanical modelling,
contributed to a wave of marathon world records. World Athletics ultimately
introduced regulations capping stack height and restricting plate
configurations — a governing body intervening to limit the
performance contribution of technology after the fact.
Similar dynamics are emerging in swimming (suits), cycling
(aerodynamic positioning systems), and alpine skiing (equipment geometry). In
each case, AI is accelerating the pace at which equipment design can be
optimised, and governing bodies are perpetually behind the technical
frontier. The World
Athletics technical rules represent one attempt to manage this, but
enforcement at the frontier of AI-designed materials science is genuinely
difficult.
Prosthetics and the Boundary of the Human
Athlete
Perhaps the sharpest question AI raises in Olympic sport concerns
the integration of prosthetics and neural interfaces. Blade runners have
competed at able-bodied sprint events for years; brain-computer interfaces
are advancing toward the point where athletes with neural augmentation could
compete in cognitive and physical disciplines.
The International Paralympic Committee and the IOC both face a
definitional crisis: what is the boundary of human performance? If an
AI-optimised prosthetic confers a measurable advantage over biological limbs
in certain conditions, in which classification should that athlete compete? These
questions are not hypothetical; they are already live in para-athletics
governance and will become more complex as neural AI augmentation
matures.
The broader conversation about where algorithmic systems end and
human agency begins — explored in the context of AI
and identity — is perhaps most legible in sport, where the
line between enhancement and replacement is drawn in competition rules that
struggle to keep pace with what technology makes possible.
Sport has always been a mirror for the values a society uses to
define fairness, effort, and achievement. AI is not breaking that mirror. It
is showing a reflection that is harder to look at without asking difficult
questions about what, exactly, we are celebrating when we celebrate an
Olympic champion.
The question of who benefits from AI-optimised athletic
performance connects to the
wider digital divide between well-resourced and under-resourced
participants in any AI-augmented domain. In sport, that divide plays out in
medals tables rather than income distributions, but the structural dynamic is
identical.
The mental health dimension of AI in sport is
less frequently discussed than the performance optimisation angle, but is
equally significant. Athletes who train with AI performance systems receive
continuous feedback that can be granular to the point of psychological
strain. Every micro-inefficiency in technique is identified and quantified;
every deviation from the optimised training load is flagged. Research from
sports psychology has noted a correlation between the introduction of
high-frequency AI performance feedback and increased rates of training
anxiety and perfectionism among elite athletes. The tool that improves
physical performance may simultaneously complicate the mental environment in which
that performance must be delivered.
The sporting world is
not unique in struggling to define the boundary between human performance and
technological augmentation, but it is perhaps the context where that boundary
is most consequential and most publicly contested. Every governance decision
about AI in sport sends a signal about what we believe sport is for: the
celebration of unaugmented human achievement, or the pursuit of optimal
performance by whatever means technology makes available. That question does
not have a single correct answer, but it deserves a more deliberate
collective response than the reactive rule-making that currently
characterises most sports governance.
The equity
implications of AI-optimised athletic training connect directly to the
broader question of who benefits when AI-driven infrastructure is deployed
unevenly across well-resourced and under-resourced participants in
any competitive system.
About the
Author
Stuart Kerr is a technology correspondent at
LiveAIWire, covering artificial intelligence, emerging technologies, and
their impact on society and industry.