Bu çalışma Asya Hun, Göktürk ve Uygur dönemlerinde Türk toplumlarının kutsal mekân anlayışlarının geçirdiği değişim ve dönüşümü literatür verilerden elde edilen bilgilerle metin girdileri oluşturulmasına ve metin girdileri yardımıyla üretken yapay zekada görselleştirilmesi kurgusuna sahiptir. Çalışma kapsamında öncelikle Asya Hun, Göktürk ve Uygur dönemlerine ait kutsal mekân kullanım ve anlayışlarına dair literatür taraması gerçekleştirilmiştir. Elde edilen veriler ile her dönem için metin girdileri oluşturulmuştur. Oluşturulan metin girdileri Midjourney isimli üretken yapay zekâ desteği ile görselleştirilmiştir. Her döneme ait 6 adet görsel elde edilmiştir. Ardından bu görseller dönemsellik, mekân tipi, mekânsal organizasyon ve somut öge kullanımı kapsamında değerlendirilmiştir. Böylece hem yapay zekâ destekli oluşturulan görseller hem de görselleri üretmek için oluşturulan metin girdileri değerlendirmeye alınmıştır. Bulgular, üç dönem için oluşturulan görsellerde dönemselliğin genel olarak sağlandığını; ancak kimi tarihsel detayların farklı dönemlere ait unsurlarla karışabildiğini göstermektedir. Mekân tipi ve mekânsal kurgu çoğunlukla başarılı biçimde elde edilmekle birlikte, özellikle Göktürk dönemine ait iç mekân kurgularında metinsel girdiler güçlendirilse dahi istenen iç mekân organizasyonlarının üretilemediği görülmüştür. Somut ögeler görsellerde büyük ölçüde yer alsa da bu ögelerin boyut ve konum doğruluğunda sapmalar oluştuğu saptanmıştır. Sonuç olarak çalışma, üretken yapay zekânın tarihsel kutsal mekânların kavramsal düzeyde temsilinde güçlü bir araç olduğunu; ancak tarihsel doğruluk ve dönemsel ayrışmayı artırmak için metin girdilerinin daha detaylı, açık sınırlarla tanımlanmış, akıcı ve aşamalı biçimde kurgulanması gerektiğini ortaya koymaktadır. Ayrıca gelecek çalışmalarda uzman doğrulaması ve farklı araçlarla karşılaştırma gibi değerlendirme çerçeveleriyle çıktının bilimsel güvenilirliğinin artırılması önerilmektedir.
This study is designed around generating text prompts based on information obtained from the literature on the transformation and change in the conceptions of sacred space among Turkic societies during the Asian Hun, Göktürk, and Uygur periods, and subsequently visualizing these prompts through generative artificial intelligence. Within the scope of the study, a literature review was first conducted regarding the use and understanding of sacred spaces in the Asian Hun, Göktürk, and Uygur periods. Based on the data obtained, text prompts were created for each period. The generated text prompts were visualized with the support of a generative artificial intelligence tool called Midjourney. Six images were obtained for each period. These images were then evaluated in terms of periodization, spatial type, spatial organization, and the use of tangible elements. Thus, both the AI-generated images and the text prompts created to produce the images were included in the evaluation. The findings indicate that periodization was generally achieved in the images produced for the three periods; however, certain historical details could become intermingled with elements belonging to different periods. While the spatial type and spatial configuration were obtained successfully in most cases, it was observed that the desired interior spatial organizations could not be generated, particularly in the interior compositions of the Göktürk period, even when the textual inputs were strengthened. Although tangible elements were largely present in the images, deviations were identified in the accuracy of these elements in terms of their size and placement. Consequently, the study demonstrates that generative artificial intelligence is a powerful tool for representing historical sacred spaces at a conceptual level; however, it also reveals that, in order to enhance historical accuracy and period-specific differentiation, text prompts should be constructed in a more detailed manner, with clearer boundaries, and in a fluent and incremental structure. In addition, for future studies, it is recommended to increase the scientific reliability of the out puts through evaluation frameworks such as expert validation and comparisons across different tools.