Objective: As the global livestock sector shifts from periodic manual inspections toward continuous objective assessments, it is undergoing a digital transformation driven by Precision Livestock Farming (PLF). This review aims to evaluate automated monitoring systems for beef and dairy cattle welfare.
Methods: This review synthesizes evidence from 118 publications, identified through a systematic search of five major databases. We critically assess two primary technological categories: non-contact computer vision systems and wearable Internet of Things (IoT) infrastructures. The computational evolution driving these tools is also evaluated, mapping the transition from traditional statistical methods to advanced deep learning architectures, such as YOLO variants and Vision Transformers (ViTs). Key Findings: Current evidence demonstrates that these systems can effectively monitor ingestive dynamics, detect early locomotion anomalies, and assess micro-environmental comfort. However, a principal limitation in the existing literature is the frequent lack of rigorous external validation across heterogeneous commercial farms, which complicates the translation of high algorithmic accuracy into robust biological validity. Conclusions: While these technologies hold significant potential to improve cattle welfare, they also introduce socio-technical risks, including data objectification and reliance on unvalidated alerts. Recommendations: This review highlights the critical need for standardized validation protocols and discusses Generative AI (GenAI) and Large Language Models (LLMs) strictly as emerging, prospective research directions rather than currently established PLF solutions.

Objective: As the global livestock sector shifts from periodic manual inspections toward continuous objective assessments, it is undergoing a digital transformation driven by Precision Livestock Farming (PLF). This review aims to evaluate automated monitoring systems for beef and dairy cattle welfare.
Methods: This review synthesizes evidence from 118 publications, identified through a systematic search of five major databases. We critically assess two primary technological categories: non-contact computer vision systems and wearable Internet of Things (IoT) infrastructures. The computational evolution driving these tools is also evaluated, mapping the transition from traditional statistical methods to advanced deep learning architectures, such as YOLO variants and Vision Transformers (ViTs). Key Findings: Current evidence demonstrates that these systems can effectively monitor ingestive dynamics, detect early locomotion anomalies, and assess micro-environmental comfort. However, a principal limitation in the existing literature is the frequent lack of rigorous external validation across heterogeneous commercial farms, which complicates the translation of high algorithmic accuracy into robust biological validity. Conclusions: While these technologies hold significant potential to improve cattle welfare, they also introduce socio-technical risks, including data objectification and reliance on unvalidated alerts. Recommendations: This review highlights the critical need for standardized validation protocols and discusses Generative AI (GenAI) and Large Language Models (LLMs) strictly as emerging, prospective research directions rather than currently established PLF solutions.