This article examines how algorithmic classification systems participate in the production of meta-identities, understood as operational classificatory constructs that mediate the visibility, circulation, and interpretation of digital content and its authors. The study employs a mixed-methods design combining controlled analytical simulation with qualitative interpretive analysis, systematic thematic coding, and comparative statistical procedures. Empirical data are derived from the analysis of 150 audiovisual works produced in formative workshops and interpreted by four types of agents: authors, peers, specialized human analysts, and two Large Language Model-based AI systems (ChatGPT and Gemini). Interpretations were analyzed across micro, meso, and macro levels, using a consolidated system of thematic categories with hierarchical weighting and normalization procedures to ensure inter-agent comparability. The results demonstrate a systematic and structural divergence between human and algorithmic classifications. While human agents preserve semantic plurality and contextual anchoring, AI systems tend to reorganize thematic hierarchies through semantic aggregation and stabilization, thereby privileging broad, reusable categories. This process produces recurring, opaque classificatory patterns that serve as infrastructural references for subsequent algorithmic decisions. The article contributes methodologically by offering a replicable framework for comparing human and algorithmic regimes of meaning production in digital environments.
Ferreira et al. (Mon,) studied this question.