Many people have died with COVID-19, a severely contagious respiratory disease brought on by the SARS virus, in several countries. Computed tomography (CT) and chest X-ray (CXR) were frequently used to achieve an accurate and timely diagnosis of COVID-19. However, because of the time commitment and high chance of human error, manually diagnosing the infection by radiographic imaging was very difficult. Emerging artificial intelligence (AI) approaches provide effective and accurate COVID-19 detection for developing precise and automated tools to stop the rapid spread of COVID-19. To help the researcher go in new directions, this research examines AI models currently in use for COVID-19 detection. The goal is to provide a brief meta-analysis to assist researchers in investigating potential areas of AI for future development. More than 200 research papers were gathered from reputed publishers. After these research papers underwent several rounds of review, 50 research articles were chosen. The key findings such as the AI techniques employed, preprocessing techniques, the datasets used, and the outcomes in terms of performance metrics are summarized while identifying recent advancements and trends in COVID-19 detection and classification leveraging AI. Additionally, it highlights the limitations of the cutting-edge research that need to be resolved for COVID-19 detection such as the overfitting problem of ML models. We also provide a list of potential future avenues for COVID-19 image classification research as we wrap up our investigation.