<p>Apples are one of the most highly produced fruits in the world. Accurate grading of apples not only meets the diverse demands of consumers but also maximizes sales benefits. Due to inefficiency and high error rates, traditional manual grading methods have been gradually replaced by automated grading using sensing technology combined with deep learning (DL) algorithms. This review synthesizes apple grading standards around the world and recent advances in apple grading using DL. Most standards categorize apples into 3–4 grades based on their internal and external quality attributes. Therefore, apples are priced according to the different quality grades, to meet the different consumer needs. DL algorithms, especially convolutional neural networks (CNN), have significantly improved the speed and accuracy of apple grading. The ability of DL to handle complex and large datasets enables the application of multi-view digital and spectral imaging techniques. Additionally, lightweight DL models can potentially reduce reliance on specialized detection hardware during grading. Future research might focus on creating extensive apple databases, utilizing transfer learning to enhance model robustness, and continuously refining DL models using real-time data from grading equipment.</p>