Air pollution is a type of environmental challenges that considered one of the crucial problems confronting human life in urban areas. This problem has been resulted from the abundance of automobiles, emissions from industrial production, combustion of petroleum products for the transportation and electricity generation. This paper will be considering all studies conducted recently to detect and predict the air pollution in smart cities. With the rapid development of deep learning technologies and their usage in almost all aspects of life, it has become possible to predict air quality in smart cities using deep learning techniques. This paper investigates the studies related to deep learning techniques under the framework of smart cities. This paper concluding that the majority of the researchers have a tendency to use more sophisticated and intelligent methods to handle the problem of the air pollution detection in early stages while few researches incorporated the simple techniques. The main predicted pollutants were PM2.5.