1.    Introduction
Sports injuries are a major concern because performance and injury are directly related. Injuries can negatively impact an athlete’s ability to perform, while poor performance can also increase the risk of injury (1). Injuries often result in missed training sessions and games, reducing overall availability and participation (2). Moreover, injuries can cause physiological and psychological consequences such as anxiety, fear of re-injury, and reduced motivation, all of which affect an athlete’s mental state, movement, strength, and endurance, ultimately impairing their ability to execute skills and perform effectively (3–4). Athletes with high performance demands and a history of intense training stress may be at greater risk for overuse injuries due to repetitive loading, poor movement patterns, and pain that can be associated with performance inefficiencies (5–7).
High-impact sports such as volleyball, handball, football, and basketball are among the most common contributors to overuse injuries, impaired motor control, and reduced performance in adolescent athletes, particularly when training programs are not personalized (8–9). While traditional group-based training methods provide benefits for young athletes, they may fail to account for biomechanical and neuromuscular differences or asymmetries (10–11). These asymmetries include differences in strength, power, or movement patterns between the left and right sides of the body or between different muscle groups (12). Therefore, integrating targeted neuromuscular training alongside traditional methods is crucial for optimizing athletic development and preventing injuries in youth populations (9). Identifying and addressing these asymmetries can help athletes improve performance across different sports, experience fewer injuries, and develop more balanced and efficient movement patterns (12).
In the modern era, Artificial Intelligence (AI) is increasingly integrated into sports, just as it is in many other aspects of human life. AI is revolutionizing sports training by enabling highly personalized programs based on extensive athlete data. Wearable sensors, movement analysis, and historical performance data can be analyzed by AI algorithms to tailor training regimens, optimize performance, and reduce injury risks (13–14). AI-driven approaches inspire personalized sports training by leveraging such data to create adaptive programs (15). By analyzing real-time inputs from athletes, AI-guided systems provide immediate feedback and adjustments to training regimens, helping to minimize injury risk and maximize performance (16–17). Traditional training methods in youth sports often rely on generalized programs designed for groups rather than individuals. Such approaches may overlook variations in biomechanics, neuromuscular control, and developmental stages among athletes (12,18). This one-size-fits-all strategy can fail to address asymmetries, imbalances, or individual needs, limiting performance gains and increasing the likelihood of overuse injuries (8). The principle of individual differences highlights that each athlete responds uniquely to training based on factors such as genetics, growth rate, prior experience, and physical maturity (9). Therefore, personalized training that considers these characteristics is critical for improving outcomes and reducing injury risks in adolescents (10,18). 
Although AI is increasingly employed in elite sports for training and performance analysis, research specifically examining its effectiveness in youth populations is limited. Despite the growing adoption of AI-powered tools, their application and evaluation in adolescent athletes remain relatively underexplored. A research gap exists regarding whether AI-guided systems provide measurable benefits over conventional training methods for young athletes across multiple team sports.
Football is a high-intensity, multidirectional sport that demands speed, agility, jumping ability, and functional movement quality. Therefore, the present study specifically focused on adolescent football players (19). Given the high prevalence of overuse and acute injuries in youth football, this population is particularly relevant for evaluating the effectiveness of AI-guided training. By targeting adolescent football athletes, this research addresses a critical gap in the literature regarding personalized, data-driven training in a sport where performance and injury risk are closely interconnected. Based on the literature and proposed design, the hypothesis of the present research is that AI-guided athletes will demonstrate greater improvements in FMS scores, sprint speed, agility, and countermovement jump height compared with athletes in traditional training programs. 
2.    Methods
2.1. Study Design
The present study employed a computer-generated randomization sequence (simple randomization) over a 12-week period to compare the effects of an AI-guided personalized training program with traditional training on injury incidence and performance metrics in adolescent athletes. Baseline assessments included key performance measures and injury history. Allocation concealment was ensured by an independent researcher who prepared sealed, opaque envelopes to assign athletes to groups, thereby minimizing selection bias.
2.2. Participants
A priori power analysis was conducted using G*Power 3.1 software to determine the required sample size. Assuming a large effect size (Cohen’s d = 0.8), an alpha level of 0.05, and a statistical power of 0.80 for independent-samples t-tests, the analysis indicated a minimum of 26 participants per group. To account for potential dropouts, 30 participants were recruited for each group, resulting in a total sample size of 60 athletes. All participants were residents of a football academy, where training is conducted daily throughout the year. The athletes were between 14 and 17 years old, lived in academy dormitories, and followed similar daily routines. Eligibility criteria required a minimum of two years of football training experience, active participation in competitive youth leagues, and no injury within the previous three months. Athletes were excluded if they had chronic illnesses limiting participation, existing musculoskeletal disorders, or used personal performance-monitoring systems. Before data collection, informed consent was obtained from all participants and their parents. Institutional ethical approval was also secured from the first author’s institution.
2.3. Procedure of Data Collection
Athletes in the AI-guided training group (experimental group) followed a personalized training program designed through AI-based software that incorporated machine learning analysis and wearable sensor data. The system adjusted weekly training loads, mobility drills, strength and conditioning, and recovery timing based on heart rate variability, real-time movement feedback, fatigue index, and Functional Movement Screen (FMS) scores (Table 1). The AI-guided group completed three sessions per week (60–75 minutes per session) for 12 weeks. Each session included a warm-up (10–15 minutes), strength and conditioning (25–30 minutes), agility and sprint training (15–20 minutes), and recovery and flexibility (10 minutes). In parallel, the traditional training group (control group) trained with the same frequency (three sessions per week) and followed a training schedule provided by certified coaches. Their program included general and specific warm-ups, agility and sprint training, power training, and plyometric exercises that are standard for the sport.

2.3.1. Functional Movement Screen
Before conducting the Functional Movement Screen, the researcher measured each participant’s tibial length (from the top center of the tibia to the floor) and hand length (from the tip of the longest digit to the distal wrist crease). Participants performed each of the seven fundamental movement patterns three times: deep squat, hurdle step, in-line lunge, shoulder mobility, rotary stability, active straight leg raise, and trunk stability push-up (20). A one-minute rest interval was provided between each movement pattern. Scoring was conducted by a certified FMS practitioner with two years of relevant experience, ensuring consistency and reliability across participants (21).
2.3.2. Agility
Agility tests measure an athlete’s acceleration, deceleration, and balance control. We used Stagnoli photocells (Italy) for the T-test. Each participant completed two trials, and performance was recorded in seconds. The full procedure was followed according to previously published research (20–23).
2.3.3. Speed
A 20 m area was marked for the speed test. Each participant performed two trials with a four-minute interval between attempts. The same photocells were used for both the speed and agility tests. During the speed test, photocells were fixed at a height of 120 cm and measured top speed in seconds.
2.3.4. Countermovement Jump
Initially, all participants stood upright with shoulders level and arms comfortably at their sides, with resistance bands placed around the knees and hips. They then performed vertical jumps. Each participant attempted three jumps, and the best two were analyzed (23).
2.3.5. Injury Incidence
During the entire training period, injury incidence was monitored, and the nature, location, and severity of any complaints were documented by a certified physiotherapist based permanently at the academy. All athletes were assessed before and after the training period at each session. An injury was recorded when an athlete reported pain or discomfort resulting from training or competition and missed a scheduled training session or game due to this issue. All reported injuries were categorized by type and rate, but for analysis, all injuries were converted into percentages. This monitoring process ensured early intervention and equal observation of both groups for injury-related outcomes.
2.4. Statistical Analysis
Data were organized using Microsoft Excel, and descriptive statistics were calculated as mean ± standard deviation (SD). Each group consisted of 30 participants. The normality of the data distribution was assessed using the Shapiro–Wilk test. Statistical analyses were performed using SPSS software (Version 25; IBM Corp., Armonk, NY, USA), with the significance level set at p < 0.05. Inferential statistics included paired t-tests for within-group comparisons (pre- vs. post-test), independent t-tests for between-group comparisons (AI vs. control group), and the Chi-square test for comparing injury incidence. Effect sizes were calculated using Cohen’s d and interpreted as small (0.20–0.49), medium (0.50–0.79), and large (≥0.80).
3.    Results
No significant baseline differences were observed between the AI-guided (experimental) and traditional (control) groups in age, sports experience, anthropometric characteristics, or BMI (Table 2).

The pre- and post-test results for the experimental (AI) and control groups across the measured performance variables, along with the corresponding p-values and effect sizes, are presented in Table 3. Fig. 1 graphically illustrates the pre- and post-test performance comparisons between the experimental and control groups for FMS scores, 20 m sprint time, agility test time, and CMJ height.

 

In Table 4, the experimental group of athletes demonstrated greater improvements in functional movement (FMS +20%), sprint speed (−4.93%), agility (−6.48%), and CMJ height (+11.86%). These performance gains were accompanied by a significantly lower injury incidence (10% vs. 36.7%), with a risk ratio of 3.37 and a number needed to treat (NNT) of just four athletes to prevent one injury. Table 4 presents the percentage changes and Cohen’s d effect sizes (post-test) for between-group comparisons, while Fig. 2 illustrates the superior performance gains in the experimental group compared with the control group of adolescent athletes.

 

4.    Discussion
The present study provides novel evidence that AI-guided training can enhance performance and reduce injury incidence in adolescent football players. The results confirmed significant improvements in functional movement, sprint performance, agility, and jump ability. These findings support our hypothesis and align with previous studies that highlight the effectiveness of individualized, data-driven training interventions in improving movement quality and neuromuscular efficiency (15, 24–26).
The experimental group demonstrated significantly greater improvements in FMS scores, including mobility, motor control, and movement symmetry. These improvements are particularly important during adolescence, a developmental period characterized by rapid growth and an increased risk of movement dysfunction and injury (8). A key observation of this study is the superior FMS improvement in the AI group, which is consistent with prior research showing that individualized neuromuscular training can correct asymmetries and enhance movement efficiency, thereby lowering injury risk (9, 12). Unlike coach-led training, the AI protocol dynamically adjusted training loads based on fatigue and asymmetry indices, which may explain the larger observed effect sizes. Recent systematic reviews have also emphasized the importance of individualized approaches for injury prevention in youth athletes (6, 8).
Improvements in sprint and agility performance are also noteworthy. Consistent with the findings of Alexe et al. (2024), who reported associations between functional movement quality and speed/agility in elite youth football players, our results suggest that AI-personalized training can accelerate these adaptations by optimizing neuromuscular load. The magnitude of improvement observed here may be attributed to the dynamic adjustment of training intensity informed by wearable sensor data. The AI system likely enhances neural efficiency by tailoring training intensity based on biomechanical responses and fatigue monitoring (10). This approach also helps optimize training load, prevent overtraining, and improve neural adaptations, contributing to better performance and recovery (27–28).
The experimental group of athletes demonstrated greater improvements in functional movement (FMS +20%), sprint speed (−4.93%), agility (−6.48%), and CMJ height (+11.86%). These gains were accompanied by a significantly lower injury incidence (10% vs. 36.7%), with a risk ratio of 3.37 and a number needed to treat (NNT) of just four athletes to prevent one injury. The reduction in injury incidence (10% vs. 36.7%) is particularly noteworthy. This finding supports emerging evidence that AI- and wearable-based monitoring can reduce overuse injuries by detecting early markers of fatigue and movement compensations (15, 25). The NNT of approximately four underscores strong practical value: for every four athletes trained with AI guidance, one injury is prevented. Such evidence has important implications for youth academies, where player availability is crucial and minimizing injury risk is a top priority (29). This approach included monitoring movement quality, fatigue index, and heart rate variability to tailor training adjustments for optimal performance and injury prevention (30–31). These findings further emphasize the concern that non-individualized training programs may enhance physical development and performance metrics but overlook underlying asymmetries or movement inefficiencies (32).
While the results are promising, several limitations should be acknowledged. First, the sample size was modest and drawn from a single academy, which may limit generalizability. Second, the follow-up period was restricted to 12 weeks, making it uncertain whether the improvements can be sustained long term. Third, although the AI system provided individualized load adjustments, adherence and qualitative feedback from players were not systematically assessed. Future research should evaluate AI-guided interventions across multiple sports, with larger sample sizes, extended intervention durations, and mixed-method designs that incorporate athlete and coach perspectives. Comparative cost-effectiveness analyses may also help determine the feasibility of large-scale implementation in youth sport academies. Overall, this study adds to the growing body of evidence that AI-supported training can yield superior outcomes compared with traditional methods, reinforcing the importance of adopting personalized, data-driven approaches in the development of youth athletes. 
5.    Conclusion
The key finding of this study is that AI-guided, personalized training programs are more effective for adolescent athletes than traditional training methods. Athletes in the experimental group demonstrated superior movement quality, physical performance, and injury prevention compared with those in the control group. With notable improvements in sprint speed, agility, jump height, and a marked reduction in injury incidence, AI-powered training systems represent a promising strategy for youth sports development.

Ethical Considerations
Compliance with ethical guidelines
There were no ethical considerations to be addressed in this research.
Funding
This research did not receive any financial support from government, private, or non-profit organizations.
Authors' contributions
All authors contributed equally to preparing the article.
Conflicts of interest
The authors declare that there are no conflicts of interest associated with this article.