The ultrasonic technique is one of the most widely used techniques in Non-destructive Testing to detect damage in civil engineering structures. It is a popular method of Non-destructive Evaluation (NDE) due to its versatility, high sensitivity on most materials, and ability to extract information about the location and type of defects. The applications of Deep Learning (DL) algorithms in ultrasonic inspection attracted much attention in recent years due to their superior ability to recognize damage and flaws in civil engineering structures. The initial section of the paper introduces the DL algorithms employed in the ultrasonic crack analysis. It then proceeds to examine the latest developments in autonomous ultrasonic NDE facilitated by DL techniques. Moreover, the paper also reviews the applications of DL to defect-related problems, such as defect detection and classification. This review paper focuses on exploring the use of DL techniques for ultrasonic crack characterization, with an emphasis on quantifying the associated uncertainty.