This paper focuses on cortisol levels in different dog matrices, emphasising the importance of non-invasive sampling for animal welfare and research ethics. The study reviewed over 18,000 articles from various databases, identifying 78 relevant studies that met the inclusion criteria. The findings reveal a growing trend for non-invasive matrices like saliva, hair, and urine, reliably reflecting acute and chronic stress responses while minimising animal stress during sampling. Other sample types, such as blood, are becoming less and less used due to their stress-inducing factors. Such an approach aligns with the growing concerns regarding animals' welfare during research design. Each sample type contributes a unique perspective, and the matrix choice depends on the research objectives. Rather than one being universally superior or more reliable, they offer complementary data that enhance our understanding of the animal's stress profile. Cortisol concentrations may be influenced by many factors, from environmental ones, such as living conditions, to genetic and social ones. This analysis focuses on the challenges often met during cortisol concentrations research, such as small sample sizes and lack of official assessment protocols. In the future, researchers should focus on an interdisciplinary approach by incorporating behavioural measures, other physiological indicators, and refining non-invasive methods. Those practices would enhance the understanding of canine welfare evaluation and promote a more ethical approach in clinical and practical settings.