This study investigates the effect of collaborative writing in digital environments on the lexical and syntactic proficiency of French learners. Twenty undergraduates wrote narrative texts individually and then in pairs using MeetingWords, yielding a corpus of thirty texts for direct comparison. Leveraging NLP techniques, we analyzed twelve linguistic metrics through graphical and statistical methods to capture both dyad-level and global stylistic patterns. Results show that collaborative texts are not mere composites of individual outputs but rather the result of diverse processes of interaction, adaptation, and stylistic blending. While average performance remains stable, collaboration reduces stylistic variability and strengthens internal correlations among linguistic features, indicating a homogenizing effect. These findings highlight the value of lexical and syntactic metrics for formative assessment and learner modeling. The proposed visualization tools—such as radar charts and PCA plots—offer actionable insights to support feedback, guide instruction, and promote reflective language learning in collaborative settings.