Most of the attempts to develop and validate tools for the automatic assessment of Critical Thinking (CT) related-skills applied Natural Language Processing techniques (NLP) to English written texts, with a few applications in other languages. Therefore, this research was aimed at understanding which NLP features correlates with six CT sub-dimensions in essays written in Italian language. 206 Master Degree students’ pre-post essays were assessed both by human evaluators and by an algorithm which automatically calculates different kinds of NLP features. We found a positive internal reliability and a medium to high inter-coder agreement of the human evaluators. Three NLP indicators significantly correlate with CT total score: Corpus Length, Syntax Complexity, and an adapted measure of Term Frequency-Inverse Document Frequency.