Toxic analysis of social media data aims to identify harmful language, including hate speech, bullying, and abusive comments, to promote healthier online environments. This paper presents a cellular automata-based approach for toxic text classification. The results show that it has the ability to capture text context and patterns through simple, localized interactions. Our proposed model encodes text data using Unicode formats, which are used as rules for the CA evolution process. The CA signals are then used as input for machine-learning models. This hybrid approach enables the classification model to detect nuanced toxic expressions. Experimental results on benchmark datasets demonstrate that the CA-based toxic analysis model achieves comparatively good accuracy in identifying class labels of toxic and non-toxic comments.