Generative AI (GenAI), particularly large language models such as ChatGPT, is increasingly used in English language education, but the conditions supporting meaningful personalization remain unclear. This interpretivist qualitative case study examined GenAI adoption in higher-education and language-institute contexts through three stakeholder groups: educators (n = 8), adult learners (n = 12), and AI professionals (n = 3). Data comprised semi-structured narrative interviews, mixed-role focus groups, four weeks of student reflective journals, and AI-mediated artifacts. Framework Analysis in NVivo was combined with source triangulation to examine convergence and dissonance, and Critical Discourse Analysis was applied to purposively selected artifacts. GenAI supported adaptive scaffolding, real-time formative feedback, profile-sensitive task design, engagement, creative risk-taking, and critical reflection. Four constraints qualified these benefits: unequal access and AI literacy; opaque privacy and data-governance practices; over-reliance that weakened metacognitive monitoring; and inconsistent handling of cultural-pragmatic nuance and affective resonance. The findings position GenAI as a personalization engine within a human-led pedagogical framework in which teachers curate, contextualize, and ethically oversee AI use. The study supports stronger AI literacy, assessment redesign, equitable access, and transparent data governance, while highlighting the need for longitudinal evaluation of durability, transfer, and inclusivity.