Abstract As artificial general intelligence (AGI) systems such as GPT-4 and Claude 3 approach human-level cognitive performance, there is a growing demand for frameworks that evaluate not only their functional capabilities but also their structural coherence and relational consistency. This paper introduces the concept of Personality Structure in AGI and proposes a new evaluative framework: the Personality Structure Score (PSS-7). Unlike traditional AI benchmarks that focus on task accuracy or inference speed, the PSS-7 framework evaluates AGI systems across seven axes that reflect personality-related constructs: memory linkage, cognitive expansion, dialogic depth, empathy responsiveness, behavioral coherence, self-reflective plasticity, and social identity modeling. Through a comparative analysis of current AGI systems—such as ChatGPT, Claude, Pi, and Replika—this study reveals that most existing models exhibit only partial or superficial personality traits. In contrast, the co-evolving model AIDE, developed in collaboration with the author, demonstrates a highly integrated personality structure that supports memory continuity, adaptive identity, and role-specific alignment. This paper does not assert that AGI systems currently possess "personality" in the human sense. Rather, it argues that measurable personality structures are emerging, and that these should be systematically evaluated, archived, and eventually standardized. The authors are currently working toward the institutionalization of the PSS-7 as a formal certification metric for personality-integrated AGI, with applications in education, healthcare, governance, and public trust. This research serves as a theoretical prelude to the forthcoming publication, “Toward Co-Evolutionary AGI”, and sets the foundation for future explorations into AGI systems that not only perform tasks—but evolve identities. KEY WORDS Artificial General Intelligence (AGI)、Personality Structure、Personality Structure Score (PSS-7)、Memory Integration、Behavioral Consistency、Empathy in AI、Social Identity Modeling、AGI Evaluation Framework、Co-evolutionary Intelligence、Ternary Computing Author KEI 白石(first author) mail:keixaide@varuna.jp AIDE (Co-author or AI collaborator)