The rapid spread of fake news across digital platforms has made automated detection an essential task in combating misinformation. In this paper, we investigate the use of cellular automata (CA) as a novel feature extractor for text-based fake news detection. CA, known for modeling complex patterns through local interactions, is used to transform textual data into structured feature vectors. These CA-derived features are then input into both traditional machine learning and deep learning models for classification. Our goal is to enhance the ability of these models to detect subtle linguistic and contextual patterns that may be missed by standard NLP techniques. This study focuses solely on textual data and evaluates the performance of CA-based feature extraction against commonly used vectorization methods, such as Count Vectorizer, TF-IDF, and Hashing Vectorizer. By comparing classification results across these different feature representations, we aim to assess the effectiveness and uniqueness of CA in capturing informative characteristics from text. Future work will extend this methodology to multimodal data, including images, to explore broader applications of CA in fake news detection and other classification tasks.