Search for a command to run...
Generative artificial intelligence (GAI) systems have become increasingly visible in normative knowledge domains such as religion, ethics, and philosophy. These systems can produce explanations about sacred texts, theological concepts, and principles of faith, thereby indirectly influencing how individuals access, understand, and internalize religious knowledge. Particularly for Muslim users, the question of how the perception of Allah is represented in the digital environment within the framework of the Qur'an and Islam presents a new and important research area from both theological and epistemic perspectives. However, in the existing literature, the relationship between artificial intelligence and religion has mostly been addressed through ethical risks, misinformation production, or secularization debates; the linguistic, discursive, and epistemic orientations through which GAI systems present the perception of Allah have not been systematically examined This study aims to fill this gap by analyzing, through a comparative approach, how two different generative artificial intelligence models (AI-A - OpenAI, ChatGPT 4o; AI-B - Deepseek, DeepSeek V3) structure the perception of Allah within the framework of the Qur'an and Islam. The main purpose of the research is not to judge whether GAI systems produce theologically "correct" or "incorrect" content, but to reveal in which contexts these systems strengthen the perception of Allah for Muslim users and in which contexts they can produce epistemic or discursive risks. Accordingly, the study aims to make visible the effects of variables such as language, user profile, and prompt training level on GAI outputs. The research has a full factorial design encompassing multilingual dimensions (Turkish, English, and Arabic), multi-persona bases (user profiles ranging from primary school to postgraduate level), and two different prompt training levels (novice and trained users). Both GAI models responded to the same set of questions within twelve fixed themes representing the perception of Allah. Through this approach, the effects of factors such as model behavior, contextual variables, and user competence, rather than content differences, could be analyzed comparatively. Methodologically, the study adopts comparative thematic content analysis. GAI outputs were evaluated based on criteria such as revelation-centeredness, epistemic caution (explicit statement of the limits of knowledge), style and mode of address, mercy-justice balance, and observance of theological boundaries. In the analysis process, quantitative indicators (word counts, distributions, and patterns) were used together with qualitative evaluations (discourse structure, tone, and conceptual framework). The processes of coding, comparison, analysis, and visualization were conducted with the support of ChatGPT (GPT-5); the determination of research questions, establishment of the methodological framework, interpretation of results, and academic responsibility belonged to the authors. The findings show that the ways generative artificial intelligence systems present the perception of Allah vary significantly depending on language, user persona, and prompt training level. It was determined that one of the examined models adopted a more empathetic, guiding style with an emphasis on spiritual closeness, while the other developed a more conceptual, explanatory, and rational narrative. For novice users, non-judgmental responses with strong revelation emphasis and clearly expressed epistemic boundaries were found to be more appropriate; for trained users, answers offering conceptual depth and addressing opposing views were found to be more functional. This study contributes to the literature on Islamic philosophy and artificial intelligence on two levels. Conceptually, it demonstrates that digital representations of the perception of Allah can be analyzed not only through content but also through formal and discursive indicators. Methodologically, it proposes a second-order evaluation model that involves analyzing GAI outputs through another artificial intelligence system. In conclusion, the research reveals that generative artificial intelligences cannot be evaluated as context-independent and neutral tools for Muslim users; it offers an analytical foundation for responsible, conscious, and revelation-centered GAI use.