<p>ChatGPT can generate poems; therefore, the implications for English poetry classes must be investigated in more detail. The purpose of this multiphase mixed methods research was fourfold, namely (a) to explore ChatGPT-3.5’s poem generating and justification capabilities, (b) to determine if ChatGPT-4o can perform better than ChatGPT-3.5, (c) to identify adoption driving factors, and (d) to design strategies for responsible usage of ChatGPT in English poetry classes. Therefore, elements of explorative case studies and comparative, quantitative and design-based research were integrated in an overarching multiphase mixed methods framework to answer the related research questions. ChatGPT-3.5 successfully generated a haiku, an acrostic, a sedoka, and a satirical poem. However, it could not be guided to improve the rhyme scheme of its ballad. Instead, it generated more ballads' demonstrating the same error. ChatGPT-4o performed better, as it not only improved the rhyme scheme but also persevered the haunting and lyrical aspects of ChatGPT-3.5’s ballad. ChatGPT-3.5 made valid points during justification of its poems, but it also hallucinated convincingly. Both models hallucinated examples of alliteration and assonance. The results of the artificial-intelligence-adoption-prediction-model (AIAPM) indicated that all integrated factors can drive the adoption of ChatGPT in English poetry classes. Therefore, learning-oriented strategies are suggested to promote responsible usage of ChatGPT in English poetry classes. This research serves as baseline for exploring the poem generating and justification capabilities of future models as ChatGPT-4.5 and ChatGPT-5 might be launched as soon as 2025.</p>