What is it about?
Artificial intelligence (AI) is increasingly being used to deliver personalised coaching that helps people achieve behavioural, physical, cognitive, or educational goals. However, research on these systems has grown across many different fields, making it difficult to understand how they are designed and where important gaps and opportunities remain. We conducted a systematic mapping review of personalised AI coaching technologies across healthcare, education, sport, workplace training, and other domains. By analysing 42 AI coaching systems, we examined how they personalise coaching, the AI techniques they use, how they are evaluated, and where future research is needed.
Featured Image
Why is it important?
AI coaching technologies have the potential to provide personalised support at a scale that would be difficult to achieve with human coaching alone. Yet the field has developed in a fragmented way, with different disciplines using different definitions, technologies, and evaluation approaches. This review provides the first broad synthesis focused specifically on AI-driven personalised coaching systems across domains. It offers researchers and practitioners a clearer picture of current practice, identifies methodological weaknesses, and highlights opportunities for designing more effective, transparent, theory-informed, and trustworthy AI coaching technologies. Our findings show that most personalised AI coaching systems have been developed for healthcare and healthy behaviour change, while education, workplace coaching, and other domains remain comparatively underexplored. Personalisation typically adapts coaching to individual users, but is often not grounded in behavioural theory or expert knowledge. Many studies evaluate whether coaching improves outcomes without independently assessing whether the personalisation itself contributes to those improvements. Most systems rely on mobile apps or web interfaces, with relatively few using more advanced robotics, virtual reality, or game-based environments. The review also found growing adoption of deep learning alongside symbolic AI approaches and identifies several priorities for future research, including broader application domains, hybrid AI methods that combine performance with interpretability, stronger evaluation of personalisation, and greater attention to ethical and real-world deployment considerations.
Perspectives
When I started this review, my main question was surprisingly simple: what actually makes a technology a coaching technology? Across healthcare, education, sport, and professional coaching, the term "coaching" is used differently and remains inconsistently defined. Rather than focusing on disciplinary labels, I found it more useful to look at the functional characteristics and goals of coaching, allowing similar systems to be considered together under a common conceptual framework. I was also surprised by how few education systems met this definition since typically focused on teaching or tutoring rather than facilitating self-directed development. Equally striking was the number of systems labelled as AI coaching that ultimately relied on ad hoc state machines, heuristics, or scripted logic without AI-driven reasoning for coaching decisions. I hope one contribution of this review is encouraging more rigorous classification and evaluation of personalised AI coaching technologies, where both the coaching process and the personalisation mechanisms are assessed independently rather than only through overall system outcomes.
Jonathan Vitale
Macquarie University
Read the Original
This page is a summary of: Personalised AI Coaching Technology: A Systematic Mapping Review, ACM Transactions on Interactive Intelligent Systems, June 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3820900.
You can read the full text:
Contributors
The following have contributed to this page
