This study examines the Agnostic Meaning Substrate (AMS), a proposed language-agnostic layer in large language models (LLMs) in which conceptual meaning stabilizes beyond symbolic language. AMS is presented as a testable theoretical framework suggesting that semantically equivalent inputs converge within a shared latent structure despite differences in wording, syntax, or language. The hypothesis was evaluated through 44 experiments comprising 141 tests conducted across 9 LLMs and 30 languages using paragraph–sentence comparisons, multilingual prompts, fragmented inputs, and emoji interpretation tasks. Results showed consistent positive Gestalt gain between paragraph and sentence representations, with large effect sizes (Cohen’s d ranging from 1.87 to 2.34) indicating strong semantic consolidation. Cross-lingual alignment was also observed in semantic similarity measures (mean cosine similarity of 0.651, increasing to 0.748 when excluding anomalous cases). Polyglot prompts (multilingual, including low-resource scripts) and fragmented linguistic inputs still converged toward stable semantic representations. These findings suggest that meaning in LLMs may emerge from an underlying geometric or topological substrate rather than from surface symbolic structure alone. The AMS framework therefore provides a testable perspective on semantic stability in artificial intelligence (AI), with implications for interpretability, multilingual representation, computational semantics, and the ethical development of AI systems. Received: 21 August 2025 | Revised: 27 November 2025 | Accepted: 10 April 2026 Conflicts of Interest The author declares that he has no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available in OSF at https://osf.io/bnh6u/?view_only=3e08e741c0974c8bbffc03380b914407. Author Contribution Statement Russ Palmer: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization.