Accurate methods to assess dietary intake underpin researchers’ ability to understand the impact of diet on health outcomes. Traditional dietary assessment methods, such as diet records, are labour-intensive for participants and subject to misreporting. Image-based dietary assessment, using smartphones or wearable devices, objectively record dietary intake but the need for extensive manual image analysis by trained nutritionists or dietitians restricts its scalability and widespread implementation(1). Recent advances in Artificial Intelligence (AI), have enabled the development of automated image-based dietary assessment systems which may overcome current limitations of image-based dietary assessment. This scoping review aimed to explore the AI models used in the automated identification and/or estimation of volume and/or nutritional composition of food and beverages in images. Five databases were searched: Medline, Web of Science, Embase, ACM Digital Library, and CINAHL and the JBI methodology was employed(2). Two hundred and sixty eligible studies published between January 2022 to March 2026 were identified. Most first authors were affiliated with institutions in China (n=72), followed by the United States (n=31) and India (n=27). Studies were predominately undertaken by teams composed solely of information technology professionals (n=222). AI models mostly employed Convolution Neural Networks architectures (n=180), and alternative architectures such as transformers, Large Language models and Vision Language Models were limited, with several studies exploring hybrid approaches combining multiple model types. Most studies focused on food image recognition (n=221) reporting percentage accuracies ranging between 21% to 100%. In comparison, volume estimation (n=67) and nutrition composition estimation (n=86) were less commonly recorded. Studies reporting nutrient estimation outcomes used 62 different food composition databases comprising of 17 national databases and 15 commercial databases. Despite growing interest in automated image-based dietary assessment, current research has predominately focused on recognition. AI models show considerable promise for improving the accuracy, efficiency, and scalability of dietary assessment, enabling applications in precision nutrition, chronic disease management, and population health monitoring. However, interdisciplinary collaboration between information technology and nutrition professionals is needed to enhance the validity and practical relevance of AI systems. Future directions need a greater focus on improving automated volume and nutrient estimation capabilities.