The recognition and classification of medicinal plants is a very significant task in healthcare, pharmacology, and biodiversity conservation. Traditional plant identification methods rely on expertise with lots of previous knowledge, are quite slow and prone to errors. The advent of machine learning and deep learning has transformed this domain into efficient, automated, and very accurate plant identification methods. This review paper provides an in-depth analysis of the recent developments concerning the recognition of medicinal plants with machine learning, deep learning and Transformer models. Recent years have seen the development of Transformer models for natural language processing, creating new avenues for plant recognition tasks. ViT-based models develop theoretically from the strength of global dependencies in image data and demonstrate state-of-the-art performances in plant classification tasks. This paper critically discusses the strengths of these models along with their limitations, emphasizing that large-scale annotated datasets and model interpretability are usually crucial in practical applications. We further review multimodal data integration and hybrid model development that integrates machine learning, deep learning and transformer strengths to further enhance the robustness and generalizability of any plant species recognition system. Finally, the review points out key challenges and future directions based on the fact that such advanced AI techniques bear the potential to contribute toward the sustainability use and conservation of medicinal plant species.