Communicating AI for Architectural and Interior Design: Reinterpreting Traditional Iznik Tile Compositions through AI Software for Contemporary Spaces

Abstract

Artificial intelligence (AI), which has a strong potential to assist architects in conceptual and visualization stages, has been increasingly used in the field of design and architecture. This study, focusing on the AI tools that generate images from texts and offer innovative solutions to design problems, aims to evaluate the use of AI for the reinterpretation of traditional Iznik tile patterns and colors in the context of architectural design and modern interiors. The methodology consists of four stages, which are the selection of AI tools (Copilot, DALL-E 2, DALL-E 3, Midjourney), the preparation of textual prompts for testing “çini” (tile) expression, testing of the AI tools’ perception of the concepts related to Iznik tile motifs, and the creation of prompt series. The findings of our study provide evidence that current AI tools exhibit distinct features in terms of variety, conceptualization, artistic visualization, and image production, while they are hardly equipped with the necessary conceptual background to communicate with the designers for the interpretation of the traditional Iznik tiles in contemporary architectural design. Specifically, Midjourney, which could produce historically referenced contemporary designs in response to textual expressions, was more successful than other AI tools. DALL-E 2 could not visualize the expressions concerning the placement of the Iznik tile surfaces in interior spaces but was quite inspiring in terms of the images regarding the tile pattern and color. DALL-E 3 and Copilot tools produced similar images in terms of color palette and patterns, whereas DALL-E 3 was better at visualizing spatial data. Our results reveal that AI tools still need to be developed for analyzing traditional patterns, styles, and forms for contemporary design purposes. On the other hand, AI tools can develop innovative approaches, optimize the tile production procedure, and have the potential to accelerate the design process for designers by generating new and diverse ideas.

1. Introduction

One of the unique values of Ottoman traditional Iznik tile art, which was a phenomenon that prospered and was widely applied in monuments from the 15th to 17th century, is its integration with the space that it occupies. Iznik tile surfaces were designed and planned as integral elements of architectural spaces of Ottoman monuments, like mosques, hammams, and palaces. However, today, although this traditional craft survives in various forms of contemporary tile and ceramics production, tiles are often not designed together with the space but are instead added later to already existing spaces as independent panels. This creates a two-fold design problem: i) the reinterpretation of the Iznik tile patterns and colors in contemporary graphic design and ii) the integration of the design of the ceramic compositions and surfaces into the architectural design process. The tendency of users to view the architectural tile surfaces as an “art object” that can be separately placed into an already existing space makes it even more difficult to recognize Iznik tile compositions or their contemporary reinterpretations as architectural features. Within this framework, an efficient instrument must be employed to facilitate reconciliation between the architect and the user. AI technologies fill this gap and provide potential tools for designers to communicate information to employers and consumers.

Technology has created great developments in industry and art. In this context, AI has been increasingly used in the field of design studies and architecture. Recently, there has been significant research focusing on the use of machine learning methods concerning the built environment. These methods involve generating design intent data, incorporating machine learning (ML) and AI, artificial neural networks (ANNs), and deep learning into architectural design, analyzing 2D and 3D data in generative design, applying AI and ML to sustainable living spaces, urban policies, and landscape design [1,2], integrating ML into architectural education [3,4,5,6,7,8,9,10], preserving architectural heritage [11], and utilizing ML as a tool at the intersection of art and architecture [1,2,12].

AI techniques have a strong potential to assist architects in both the conceptual and visualization stages by rapidly producing and processing images [13,14,15,16]. AI tools for generating images from text face challenges in accuracy and integrating various design aspects, which architects must address with strong and updated architectural knowledge. While students often struggle with AI commands, experienced architects are more adept at using AI for inspiration and achieving effective results [17,18]. AI programs such as DALL-E 2, Midjourney, and Stable Diffusion are widely used in research to generate images from text, but they have limitations in integrating specific design styles into their outputs within a pre-designed structure [15,17]. Tools specialized in recognizing architectural styles are turning to design parameters by generating images in a variety of styles [15]. However, they have been found to fail in generating images intended to include various references to features of local structures or cities [12]. Effective image generation depends significantly on prompt quality: longer prompts and leading words improve results, highlighting the need for extensive vocabulary to increase visual fidelity [16]. This study reviews research at the intersection of AI and architecture, focusing on how AI tools that generate images from text offer innovative solutions to design problems. The methodology is developed through a case study concerning a specific ceramic type, namely Iznik tiles. Ceramics, as an artistic expression with practical use, have a wide range of uses related to their industrial applications, such as paint, textiles, automotive products, and construction in science, as well as a usage area that includes cultural approaches [19]. We argue that AI can significantly contribute to the design and production of Iznik tiles, a form of traditional Ottoman and Turkish ceramic art known for its intricate patterns and vibrant colors. It is important to note that a deep understanding and respect for the cultural and artistic significance of this traditional art form should guide its use in contemporary spaces and designs. In this paper, we elaborate on the research question of whether AI can be a tool to enhance the creativity and innovation of designers who aim to reinterpret traditional Iznik tiles in contemporary graphic and spatial designs while sustaining the spirit of Iznik ceramics. Current research on the contemporary production of Iznik tiles develops new techniques for resolving issues in production management, user-oriented or participatory design processes, and analysis of traditional patterns for historically referenced contemporary designs. Based on these studies, we aim to research the possible uses and contribution of AI in contemporary designs reinterpreting traditional Iznik tiles and their original and unique relationships with architectural spaces.

The areas in which AI can contribute to traditional Iznik ceramic tile design are as follows: (i) the development of patterns [20], (ii) analysis of color and materials, (iii) design and customization, (iv) restoration, (v) optimization of production, and (vi) control of production or production management. We discuss in this paper that AI algorithms can help designers to generate novel ceramic tile patterns and designs that draw inspiration from the traditional Iznik tiles, especially for pattern development. Additionally, our research critically examines to what extent the current AI tools can communicate to designers and produce results proper to the design questions that they pose. By critically examining the images generated by the AI algorithms in response to the design questions related to the traditional Iznik motifs, color palettes, and compositional aspects, this paper also brings light to the underdeveloped skills of AI tools concerning historically referenced contemporary designs. This paper also questions whether AI tools can come up with novel variations and combinations of traditional Iznik tile compositions. AI tools, specifically machine learning (ML), can enhance the use of nanometer-sized pigments in tile manufacture by optimizing the examination of color and materials. This has the potential to create colors that are brighter and more consistent. This method can enhance the quality and overall aesthetic of the glazing while also aiding in the preservation of its traditional appearance [21].

Another field in which AI-powered CAD systems can help artists and designers is speeding up the design process by providing many suggestions and automating some phases of the design process. Moreover, these technologies can improve design workflows by executing intricate tasks like pattern recognition and decision making, resulting in increased efficiency and creativity [17].

Concerning the conservation of monuments and sites, a significant addition of AI is its ability to examine traditional Iznik tiles using advanced techniques. This enables AI tools to accurately differentiate between original tiles and replicas, making it a valuable tool in preserving the authenticity and uniqueness of designs, which is crucial for safeguarding cultural heritage and ensuring genuineness and distinctive design in restoration endeavors [18].

Furthermore, the integration of AI-supported automation and robotics technologies can be employed to optimize the production process. This enables manufacturers to enhance the efficiency and quality of tile production while maintaining the handcrafted quality of Iznik tiles. As a result, to achieve superior outcomes, AI can assist in identifying and recommending optimal materials and techniques [22]. In Iznik tile production, for instance, problems that may arise during the clay selection, primer use, and glazing stages directly affect the quality of the final product [23]. AI technologies can be used to conduct quality control by identifying errors or inconsistencies and promptly delivering feedback to craftspeople.

In sum, by harnessing the abilities of AI to recognize patterns, enhance productivity, and optimize processes, it is possible to enhance the design and production of contemporary ceramics and traditional Iznik ceramic tiles and to develop creative interpretations of Iznik tile patterns for contemporary architectural spaces [20]. Therefore, AI can have a crucial impact on the conservation and sustainability of Iznik tile heritage and its contemporary interpretations and production. It can improve creativity, ensure historical accuracy, and optimize manufacturing processes.

Based on the experiences of three of the authors of this paper, who have been working as designers at Iznik Tile and Ceramic Design Office and Production Atelier (İznik Çini ve Seramikleri), which is a private company based in Istanbul and Iznik in Türkiye since 2020, several observations have been made concerning the contemporary design processes of traditional Iznik tiles. By evaluating design problems faced at the company’s design office and by observing the approaches of designers and craftspeople to these problems, we classify design problems according to four fundamental dimensions of Iznik tile production: (i) material, (ii) production, (iii) economy, and (iv) aesthetics.

The company’s main aim is to reproduce traditional material, ratios, and production approaches following the results of the scientific studies on the original Ottoman-period Iznik tiles found during 16th-century Iznik kiln excavations or on the Ottoman monuments that contain original Iznik tiles. An example of the company’s claim of being able to produce “original” tiles is one of the tile products, which is 23.5 cm × 23.5 cm in dimension. This specific dimension of tiles was repeatedly used in the 15th- to 16th-century Ottoman monuments. While creating a reliable brand for accessing originally reproduced Iznik tiles for users, this situation also restricts flexibility in the production process, determining—as explained above—the minimum and largest tile sizes. As a result, the high costs of producing new molds in different sizes make designers and users abandon project-specific dimension proposals. Such issues directly affect the aesthetic dimension and design.

Aesthetic problems may also arise from user expectations and producer guidance. Firstly, aesthetic problems may result from the users perceiving tiles only as historical elements. We believe that reproducing traditional designs either directly or with minimal adjustments in pattern and composition limits the contemporary designers’ contribution to the Iznik tiles’ heritage and sustainability. Instead of the historicist approach, which is directly imitating traditional designs, there should be an effort to adopt a design approach that creates contemporary designs by referencing the original Iznik tiles’ tangible and intangible cultural heritage.

Secondly, aesthetic problems come up because of the designers and users who consider tile design separately from architectural design and spatial data. As referred to above, traditional Ottoman Iznik tiles were always designed with strong relations to space organization and other architectural features of space. Thus, the problematic of reinterpreting the relationship between space and tiles is central to the discussion of contemporary designs inspired by Iznik tiles.

In this context, the main research problem of this manuscript is to discuss the areas in which AI can contribute to the design and production process of tiles, especially Iznik tiles. The Iznik tile tradition is a distinctive form of Ottoman ceramic tile tradition that is known for its unique and meticulous production process. This outstanding traditional craft was listed in UNESCO’s Representative List of the Intangible Cultural Heritage of Humanity and was influential in the inclusion of the city of Bursa, which is the province that Iznik depends on administratively, to the UNESCO Creative Cities Network.

In this context, our research questions are as follows:

• Is it possible to create contemporary graphic and interior designs that incorporate traditional tile patterns by using AI? In which stages of the design process can AI tools contribute to the architect and architectural design?
• What strategies may be developed for integrating AI tools to generate designs that represent and contribute to the cultural values of Iznik tiles’ heritage?
• What is the role of AI in enhancing traditional or historically influenced tile designs in modern environments? Do the recommendations generated by AI suffice for application in the realm of design?
• Which methods can be used to create AI-powered generative design tools that support designers in creating customized tile patterns based on traditional motifs?
• Do different tools have equal efficiency in terms of interpreting motifs, generating alternatives, or presenting design alternatives that are integrated into the space or architectural project? What are the potential results of evaluating and comparing the efficacy of different application tools or software in creating space-integrated tiles?
• In what ways can AI analyze traditional patterns and forms to suggest modern design approaches that align with the expectations and psychological needs of contemporary users?
• Can experimental methodologies determine the context in which AI recognizes tile-related and spatial terminology?

Based on these research questions, the objectives of the research are (i) to explore the potential of AI technologies in the integration of Iznik tile compositions within contemporary architectural designs as well as historically referenced architectural interiors, (ii) to explore how AI software can be communicated in order to reinterpret traditional Iznik tile compositions within contemporary spaces, (iii) to identify the stages of the design process where AI might make a substantial contribution, (iv) to develop strategies for merging AI tools that emphasize creativity with those prioritizing artistic visualization, (v) to investigate experimentally how AI can analyze traditional patterns to propose designs, (vi) to examine the effectiveness of different AI tools in interpreting motifs, generating alternatives, and integrating designs into architectural spaces, and (vii) to investigate the context in which AI recognizes tile-related and spatial terminology.

To discuss broadly the above-mentioned aims, we structure five sections following the introduction part. Firstly, we explain the characteristics of the historical Iznik tile design and production and we underline Iznik tiles’ integrity with space. Secondly, we discuss the current literature on the use of AI for architectural and design purposes by specifically prioritizing studies on AI-generated tile design. Thirdly, we discuss our experimental method incorporating AI technology, tools, and software and AI-supported Iznik tile designs produced by text-to-image software. Fourthly, we evaluate the results of our experimentations concerning the use of AI for architectural space-bounded tile design. As a result, we argue that there is a scarcity of studies on the potential and shortages of using AI-supported technologies for the contemporary interpretations of Iznik tiles, and the originality of this study stems from the fact that we propose an original methodology to examine and understand to what extend AI can perceive traditional-craft-specific terms and how it responds to design questions prepared specifically for communicating with AI to make it produce historically referenced contemporary design images.

2. Tile Design and Its Relationship with Space in History and Today

Tile design has held a significant place in various cultures and civilizations throughout history. Although tile art has evolved in different ways over time, it remains an integral part of the arts and crafts tradition. The unity of tile and architectural design was a prominent feature in 15th- to 17th-century Ottoman monumental architecture, especially in structures such as baths, mosques, educational buildings, and palaces. Consequently, Iznik tiles have become one of the defining elements of 15th- to 17th-century Ottoman architecture. Iznik tiles produced during different periods can be categorized by their colors, such as the blue-and-white period, the blue-and-turquoise period, and the polychrome period. Additionally, they can be evaluated based on criteria such as motif and craftsmanship quality [24]. There are also examples known as “Damascus style”, which, despite being found in structures in Damascus, have been identified as produced in Iznik. These tiles are technically cruder in comparison [25,26]. Two significant structures featuring tiles of this style are the Yeni Kaplıca Bath in Bursa and the Hadım İbrahim Pasha Mosque in Istanbul [27]. The technical development of Iznik tiles can be traced through religious structures such as the Süleymaniye Mosque, Rüstem Pasha Mosque, and Sokollu Mehmet Pasha Mosque, as well as various sections of the Topkapı Palace. Although various technical deteriorations in tiles began to appear by the 17th century, new experiments in the integrated use of tiles with architectural spaces were conducted, and production volumes were increased with the establishment of new production centers. During this period, tiles produced in Kütahya complemented those produced in Iznik. According to current research on Ottoman tile production between the 15th and 17th centuries, Istanbul, Iznik, and Kütahya were the primary centers for the production of tiles used in various monumental structures in the imperial cities, particularly those in Istanbul. Necipoğlu suggests that Iznik may have been in contact with Istanbul workshops from the 1530s to the 1540s [28]. Similarly, Kırımlı also highlights the contact between workshops. This observation indicates the increasing significance of Iznik tiles and suggests that, even if the tiles were produced in other cities, they were included in the Iznik tradition [29]. By the second half of the 16th century, tile art had reached a certain maturity in terms of color, pattern, and composition, and distinctive qualities related to its architectural use began to emerge [13].

Iznik craftsmen, producing in response to demands from the Ottoman Palace, enriched Ottoman architecture with their innovative designs, adding significant aesthetic value. Despite continuing their production for many years, tile production in Iznik completely ceased in the 18th century due to economic reasons. The use of Iznik tiles in palaces, inns, and baths in Istanbul, which have been preserved to the present day, has facilitated the identification and historical analysis of these tiles in research and allowed them to gain international recognition [30]. According to research on Iznik tiles used in mosques, six main points have been identified regarding the relationship between traditional architecture and Iznik tiles [13].

The Iznik tile plays a crucial role in defining the mihrab wall and narthex of sultanic mosques, making it a prominent architectural element in these structures. In general, tiles are used on the most prominent/important architectural elements. Tile compositions were placed on one or more window pediments in all mosques. Additionally, the fact that the tile compositions are usually placed in relation to windows and window rows ensures that enough light is provided for the perception and appreciation of Iznik tiles by the users.

A tile is an architectural element that provides linguistic unity between the different spaces of the mosque. The analyses revealed repetitive use of panel patterns in all mosques, except the Fatih Mosque, confirming the use of templates in the sultanic mosques that we studied. When the mosque’s exterior and interior organization are considered, it is seen that, following the symmetrical design approach of the mosque structure and space design, the layout and patterns of the tile compositions are always symmetrical. The tiles were manufactured in a manner that precisely matched the architectural elements and effectively addressed the architectural problem. Special spaces like the spaces used directly by the sultan are identified using unique tiles or colors that are not visible in other areas of the building.

Traditionally, Iznik tiles were generally rectangular or square-shaped ceramic products with floral or geometric patterns made with a black outline on a white background, and their color scheme included cobalt blue, turquoise, green, red, and purple colors [13]. The dimensions of each side of the square-shaped tiles generally varied between 21.5 cm and 25 cm. Rarely, square tiles with 30 cm sides were also found. Similarly, the short side of the rectangular-shaped tiles varied between 19 cm and 28 cm, whereas the long sides generally were around 25 cm to 33 cm, and the side length of hexagonal tiles was typically around 11 cm. Examples of dimensions of panels composed of Iznik tiles included rectangular panel dimensions of 21.5 cm × 48 cm, 74.7 cm × 24.5 cm, and 141.5 cm × 71 cm [31]. Thus, the dimensions of square and rectangular tiles were similar, while panel dimensions varied to a high extent, probably as a result of the traditional craftsmen’ and architects’ efforts to design tile panels well integrated into the interior space.

Considering the scaling of space, an average human height corresponds approximately to the middle height of the first row of windows. When the relationship between the mosque scale and the use of tiles was examined, it was seen that the tiles were mostly used at heights that users could easily see and touch, that is, around the first row of windows. In addition, the most important design principle considered in the placement of tiles is symmetry. Tile compositions were symmetrically arranged in accordance with the structure of the mosque. The placement of the tiles, the creation of the composition scheme, and the symmetry observed even in the smallest motifs provide narrative coherence to the design, demonstrating the decoration’s adherence to the architecture. With the balanced use of tiles consisting of repeating units and free compositions, the monotony brought by symmetry is prevented [32].

Based on the six design principles that form the traditional tile–space relationship, criteria for the contemporary tile–space relationship can be proposed as follows:

• Tiles can be used on the surface where it is desired to create a focus in the space.
• Tile compositions can be positioned according to windows and used in relation to rows of windows.
• In contemporary architectural design, tile compositions should be developed by taking into account the asymmetrical or symmetrical balance of the structure.
• Tile designs should be made in accordance with architectural details. Special forms and sizes can be designed to provide solutions to architectural problems.
• The hierarchy between spaces can be supported and strengthened with different tile designs.
• Tile designs are associated with human scale. Tiles can be positioned close to eye level.

Tiles found in historical buildings are considered as having cultural heritage values. In parallel with the advancing technology of 21st-century architecture, Iznik tile production and design continue to maintain their significance. Today, Iznik tile design, like other art forms, has undergone a modernization process. However, Iznik tile surfaces are rarely used in contemporary buildings and, as referred to above, are not designed integrally with the architecture. As a result, the relationships between Iznik tiles and the architecture mentioned above have not been sustained, even in a transformed manner. With technological advancements, new techniques and materials are being used in Iznik tile production. The relationship between Iznik tile design and architectural space is continuously evolving to meet modern needs and trends while preserving historical heritage.

In this process, it is crucial to consider fundamental principles such as aesthetics, functionality, and sustainability. Also, keeping up with innovations in the field of architectural design with the use of AI is another necessity today. In this context, AI tools are seen as a technology that offers potential in the design of Iznik tile together with the space, both in terms of aesthetics and function and in terms of interpretation together with the architectural characteristics of the building. In the following sections, ceramic/Iznik tile or space design approaches using AI technologies are mentioned to develop a discussion on this issue. Furthermore, staying abreast of the advancements in architectural design facilitated by AI in the present era is imperative. AI tools are seen as a technology that can enhance the design of Iznik tiles and the surrounding environment. This includes improving the aesthetic and functional aspects as well as contributing to the interpretation of the architectural characteristics of the structure. The following sections examine the use of AI technologies in developing ceramics and space design methods, particularly Iznik tile designs, to foster a comprehensive discourse on this matter.

3. Literature Review: Ceramic/Iznik Tile Design in Space and AI

In this section, the relevant literature on artificial intelligence in the fields of architecture, cultural heritage, traditional arts, and ceramics is discussed. In accordance with the unique research methodology that we developed for this study, we examined the current studies that discuss the method of generating images from text. The literature review that we conducted in this framework showed how AI technology in generating images from text affected the design sphere, especially concerning various forms of traditional and contemporary arts. Therefore, we extended the scope of the literature review by including AI studies in the fields of cultural heritage and some art disciplines.

3.1. Text-to-Image Generation in the Architectural Design Field

An artificial intelligence image generator (AI Image Generator) is a computer program based on deep learning algorithms and text-to-image conversion. The algorithm is trained on various image data and parameters so that it learns to create new images that match the user’s text descriptions. The technique of generating images in this manner is called text-to-image synthesis [13,15]. This method differs from web queries: text-to-image inquiry is typically structured with terms that describe the subject, shape, and purpose [15]. Within the scope of this study, from among the studies carried out at the intersection of AI and architecture, studies that seek innovative solutions to various design problems through artificial intelligence tools that produce images from text were examined. All the studies examined draw attention to the potential of AI as a tool that can contribute to the idea/concept phase of the architectural design process [33,34,35]. AI can access big data in short order in terms of producing images based on certain criteria and processing these images [35]. In this way, it can act as a virtual assistant to architects at the idea stage, helping to generate a wide variety of ideas and speeding up this process. It is possible to quickly obtain realistic images at the visualization stage as well as at the idea stage [14].

There are also some challenges with AI tools that create images from text. Architects, who are the users of these tools, have an important role in overcoming these difficulties. Although artificial intelligence tools are effective in creating architectural visuals, the suitability of the material and decoration details of the visuals must be checked by architects. In addition to this control, livability, structural suitability, cultural suitability, climatic conditions, and legal criteria are other qualities that require being checked [34,35]. Therefore, it is recommended that architects have sufficient/strong basic architectural knowledge and constantly update this knowledge [34]. The architect/user’s knowledge will influence success in creating text commands that AI programs can detect [15]. In the study conducted by Kwon et al. [18], it was determined that architecture students had more difficulty than professionals in recognizing the connection between design input and the final product, and therefore they created more text commands than architects. Similarly, it has been observed that, as the level of competence/expertise in the field of design increases, the architect is more open to using potential sources of inspiration and unexpected results produced by the artificial intelligence tool. Therefore, user/designer knowledge and skills gain importance in text-based visual designs to be generated in an artificial intelligence environment.

DALL-E 2, Midjourney, and Stable Diffusion, which are artificial intelligence programs trained using text–image datasets, are frequently preferred in research [33,34,36]. Although DALL-E 2 and Midjourney stand out concerning producing result images that are closer to the expectations of researchers or users, it has been stated that these artificial intelligence tools have limitations in integrating a certain design style or form into a designed structure. However, it is possible to develop new tools that can be trained through machine learning and are specialized in recognizing architectural styles [35]. When the images taken from artificial intelligence programs are examined, it is seen that artificial intelligence tools generate visuals in different styles. Thus, artificial intelligence tools turn into design parameters and actors with their unique visual expressions [37]. Some tools excel in terms of variety and creativity while others are better at artistic visualization; all tools have various limitations [36,38].

One of the most important issues to consider in the method of creating images from text is expressions, which are called prompts [35]. The language used when generating images is considered a new form of artistic expression [39]. Better quality results can be achieved with longer prompts. For example, the average prompt length for Midjourney, one of the artificial intelligence tools, consists of 27.16 words [16]. In addition to the main expressions that determine the design, the importance of guiding words is also emphasized; to improve the visual, comprehensive dictionaries of guiding words should be created regarding the space, environment, and qualities of the visual (which is the final product) [35]. Another point that should be mentioned is that these tools fail to visualize expressions loaded with metaphorical or cultural codes and to create visuals that are intended to contain various references to the features of local structures or cities [35,40].

3.2. Artificial Intelligence and Cultural Heritage: Conserving and Reinterpreting Traditional Arts and Crafts

Artificial intelligence has significant potential not only for contemporary architectural designs but also for practices regarding the conservation of cultural heritage, adaptive reuse, and sustainability of traditional arts and crafts.

The functionality of AI tools in the field of cultural heritage conservation is a current debate. AI tools can achieve a good standard in repetitive tasks, such as the identification and classification of heritage elements [41]. Similarly, they can be useful in reconstructing patterns of the damaged parts of artworks such as mosaics, stained glass, calligraphy, or cave paintings [42,43,44,45]. These studies in the field of conservation show that AI tools can be trained with previously collected and defined visual data specific to the relevant art form. In other words, a dataset can be created by introducing images specific to an art form to an AI tool [44]. However, both the current literature and our study show that tools that are not specifically developed for an art form may not be able to recognize some details and terms related to that traditional art form. For example, in the study of Moral-Andres et al. on the reconstruction of ancient mosaics, it was seen that the DALL-E AI tool could not produce complementary elements that were consistent with the whole, and DALL-E was inadequate in producing patterns except geometric ones. It also has limitations in generating correct colors [42]. However, it is anticipated that these tools will improve in the future [42,46]. Duan et al. discuss that AI technology can be used even when working on the conservation of cultural heritage on an urban scale, such as the transformation of an industrial heritage urban area [47].

Another potential of AI technology for studies in the field of cultural heritage is adaptive reuse with contemporary methods [48]. While working in this field, which means reinterpreting the values of cultural heritage in contemporary contexts, the authenticity and integrity aspects of cultural heritage should not be ignored [41]. When the current state of AI tools is evaluated, there is a risk of generating new images that do not correspond to the correct references, values, or details of the heritage items and thus convey false information or foster disinformation [41,48]. It is recommended that those working in the field of cultural heritage should be AI-literate in order to analyze and critically edit the new information and reinterpretation produced [41].

Another suggestion for integrating the unique qualities of cultural heritage into contemporary designs is to update art history and architectural design education. As pedagogical tools, AI tools can be useful in developing critical approaches to art history and architectural design for creating new interpretations of heritage [49]. Despite all the potentials of AI referred to above, it should be kept in mind that AI still has serious limitations and can be misleading in reproducing or reinterpreting traditional arts that have unique qualities and vary according to cultural codes and local and regional conditions [46,48].

3.3. Utilizing AI Technology as a Support for Designing Ceramic Products

The rapid changes in the field of design, the limited product development time, the high cost of raw materials, and the challenge of finding the most suitable model for new product design are fundamental issues in the ceramics industry. Today, AI technology is widely used in many fields. Utilizing AI technology as a support for designing ceramic products represents a new direction in the development of the ceramics industry. AI technology provides design alternatives for ceramic products, including creative design, image design, and human–computer interaction. As the era progresses, ceramic design encompasses other aspects involving functionality, economy, aesthetics, and the profound societal meanings that these elements create [50,51,52].

Among the design elements of modern ceramic products are material technology and modeling; among these, the functional factor, which is of decisive importance for material and modeling, is predominant. For instance, it has been proven that ceramic tiles produced by combining ceramic materials with new optical materials have round and transparent surfaces with clear patterns that can be altered under different light conditions. In the ceramic production process, the creator must not only possess specific skills and artistic talents but also have certain cultural achievements. Moreover, it requires the creator to work diligently and practice over an extended period. However, the complexity and highly professional nature of ceramic art creation make it difficult for non-professionals to get started. Against the backdrop of the rapid development of AI, image analysis technology has emerged, enabling non-professionals to design ceramic products. Thus, AI has provided significant creative power [53]. Traditional-style decoration has provided a solid foundation for the development of modern ceramic art. Designers need to use modern technologies, various colors, and different patterns to create their works. As a result, designers can achieve continuous innovation and change by building on traditional experiences and foundations while consistently keeping up with modern concepts [54].

The traditional ceramic-making method is composed of many stages involving a group of craftspeople with different skills. Therefore, reinterpreting a traditional piece of ceramics for contemporary design purposes and establishing a design and production plan requires numerous experiments and AI assistance. For instance, various shapes and patterns can be designed by computers, and AI technology can provide higher visual quality and artistic appeal by experimenting with texture and color options for design ideas and materials [38]. Moreover, the complexity of current product design and the limitations faced by designers in the design process necessitate the use of AI, which relies on intelligent algorithms to assist with design [20,34].

When examining the relationship between ceramic design and AI, deep learning and neural network technologies are predominantly used. The creation of the fundamental objects of traditional ceramics is being developed using generative adversarial networks (GANs), a powerful deep learning model. The significance of GANs lies in their ability to generate data significant for fields such as image creation, sound synthesis, and natural language processing [40,42]. For the research and analysis of ceramic aesthetics, studies have been conducted using methods such as convolutional neural networks (CNNs), the visual geometry group (VGG), deep learning (DL)-based computer-aided design (CAD) of ceramic images, and the preprocessing of ceramic datasets using Canny edge detection (CED) [49,50,55,56].

With the development of AI and deep learning technology, another method used in ceramic design is image style transfer. The Microsoft 2017 COCO dataset has been used for the images. The ceramic images in this dataset are sourced from the Internet and mostly consist of classic ceramic patterns [57].

Working with GANs or other image generation techniques has significantly simplified in recent years. The current literature on the issue emphasizes that user-friendly technologies such as DALL-E and Midjourney offer little opportunity for artistic expression, which diminishes their attractiveness to artists. These items tend to exploit the user; the opposite is also true. Conversely, open-source approaches provide greater artistic independence and enable a wider range of creative concepts to be explored [48]. In their studies, Guliajeva (2024) and Dixon (2023) used Midjourney and DALL-E applications to generate two-dimensional images and then used these resulting images to create three-dimensional designs [56,58].

AI technologies are beneficial tools in the fields of architecture and ceramics as they offer prompt suggestions during the initial design phase. To effectively utilize AI as a helpful assistant based on the design, the architect must possess a fundamental understanding that enables them to effectively manage and adapt to evolving advancements. This knowledge provides a potential that can be converted into design knowledge and expertise through various tests, hence enhancing the data of AI. Within this framework, the subsequent sections comprise a case study and a discussion on the relationship of text and image in relation to Iznik tiles and their spatial interpretation. This analysis is conducted within the context of the dynamic nature of AI technologies, which are often utilized in the discipline of architecture.

4. Methodology

The methodology of this paper is composed of four consecutive stages, each of which contributes to preparing the groundwork for communicating with AI for designing contemporary architectural spaces integrated with tile surfaces inspired by traditional Iznik tiles and their integration with monument architecture.

4.1. Selection of AI Tools

The authors conducted a literature survey to select appropriate AI tools for text-to-image generation for architectural and interior design purposes. The literature review revealed that tools such as DALL-E 2, Midjourney, and Stable Diffusion are frequently used in research on architectural design and artificial intelligence. Consequently, the authors decided to use the most popular text-to-image AI tool, DALL-E 2, as the starting point for this research. Following the trials with DALL-E 2, the authors experimented with Microsoft Image Creator (Copilot), which is powered by DALL-E 3′s advanced framework and is known for its accessibility and ease of use. Microsoft Copilot is a generative AI chatbot developed by Microsoft. At first, the AI program, capable of generating images from text, was powered by DALL-E 2. However, in October 2023, it transitioned to being powered by DALL-E 3. The analyses conducted within the scope of this study were carried out using Microsoft Image Creator, which is powered by DALL-E 3. To enhance variety, the authors also employed V6 Midjourney, another widely used text-to-image AI tool, known for its ability to generate high-resolution images and its effectiveness in creating large scenes due to its vast database of artistic references [59]. Therefore, the authors generated images using DALL-E 2, Copilot (Microsoft Image Creator), DALL-E 3, and Midjourney software, all based on the same prompts specifically created for architectural and interior design projects that reinterpret and incorporate traditional Iznik tiles and their relationship with contemporary spaces. Due to the rapidly evolving nature of AI technology, DALL-E 2 became outdated during the course of this study, and OpenAI ceased selling new image generation credits for it. Consequently, this study continued with DALL-E 3 instead of DALL-E 2.

4.2. Testing the Expression “Çini” (Iznik Tile)

Before developing appropriate prompts for communicating AI for historically referenced architectural and interior design, we investigated whether AI tools could recognize the term “çini” (Iznik tile) and various traditional çini motifs (Figure 1).

The term “çini” is etymologically associated with China and Chinese porcelain, and is evaluated within the scope of traditional Turkish arts, defining glazed ceramic ware and architectural pieces [60,61]. Some researchers define tiles based on the production methods and qualities of Iznik tiles, emphasizing their relationship with Ottoman art [62,63]. Since our study focuses on architectural wall ceramics, we investigated how tiles and Iznik tiles are defined in the literature.

According to this research, “çini” appears in the literature with terms such as tile, faience [64], Iznik tile [65,66], ceramic tile [67], çini, Turkish chinoiserie [64], çini-making [61,67], underglaze tile [68], and underglaze painted tile [65,66]. The most frequently repeated terms in both Turkish and English literature, namely “çini”, “Iznik tile”, and “underglaze painted tile”, were tested with the artificial intelligence tools DALL-E 2 and Copilot. For the term “çini”, DALL-E 2 did not produce meaningful results, whereas Copilot generated different visuals that matched the descriptions in the literature, including objects, utilitarian ceramics, and architectural tiles. For the term “Iznik tile”, both AI tools produced successful results. DALL-E 2 generated symmetrical tile visuals that largely matched the patterns, compositions, and colors of Iznik tiles, while Copilot predominantly used shades of blue to create densely decorated, centrally, and symmetrically arranged tile visuals.

Both AI tools converted the term “underglaze painted tile” into meaningful tile visuals; however, these visuals did not exhibit the characteristics of traditional Iznik tiles (Figure 2c,d). Based on the evaluation of 18 queries and 72 resulting visuals for the three terms, we decided to use the term “Iznik tile” within the prompts prepared for this study (Figure 2a,b).

4.3. Testing of Iznik Tile Motifs

As referred to above, Iznik tile motifs found in the Turkish and English literature were tested using AI tools such as DALL-E 2 and Copilot (Microsoft Image Creator). The Iznik tile motifs tested include rumi, hatai, rosettes, palmette, tulip, carnation, rose, spring branches, tree of life, golden horn, haliç işi, and çintemani (Figure 1). For the term “desen” (pattern), the AI was tested with both “motifs” and “pattern” to determine which yields better results. It was found that the prompts including the term “motif” produced more relevant results compared to the prompts utilizing the term “pattern”.

As a result of the experiments on the “knowledge” of AI tools concerning the traditional Iznik tile motifs, it was observed that utilized AI software did not recognize haliç işi and çintemani motifs. It was discovered that these terms that point out specific traditional Iznik tile motifs were not introduced to or were absent from the AI infrastructures used and were thus excluded from prompts.

In AI, our experiments revealed that the term “motif” gives results closer to the traditional patterns than the “pattern” term itself. It was remarkable to notice that the term “Iznik tile” yielded more favorable results with DALL-E 2 compared to the other AI applications. While we evaluated that DALL-E 2 provided results that bore the characteristics of the traditional Iznik tiles, we considered the Copilot application more successful in producing modern interpretations of the motifs, as will be discussed in the results section.

4.4. Creation of Prompt Series

To understand the knowledge of DALL-E 2 and Copilot AI tools regarding Iznik tiles and to determine the content of the prompts to be written within the scope of the study, common motifs found in Iznik tiles, such as rumi, hatai, penç, palmet, tulip, carnation, rose, spring branch, tree of life, haliç işi, and çintemani, were tested using the AI tools referred to above. The prompts entered into the AI tools were structured with the following phrases, where the blanks were filled with motif names: “Iznik tile with... motifs”, “... motifs”, “... pattern”.

The resulting images generated by AI revealed that, as is the case in haliç işi and çintemani motifs, none of the AI tools effectively recognized the rumi motif. While Copilot produced visually similar images, no proper rumi motif was detected.

On the other hand, the penç motif was recognized by both AI tools, whereas, for the hatai motif, both AIs generated visuals more similar to the penç motif.

Another result concerning the knowledge of AI concerning the traditional Iznik tile patterns is that the palmet motif was recognized by both AI tools, with Copilot producing better results.

Similarly, motifs like tulip, carnation, rose, and spring branch were successfully recognized and depicted on the tiles by both AI tools. The tree of life motif yielded meaningful results, though these did not align with classic Iznik tile patterns. As underlined above, the haliç işi and çintemani motifs produced results unrelated to traditional designs, indicating that neither AI tool recognized these motifs.

As a result, after comparatively analyzing the 264 images generated by DALL-E 2 and Copilot for 11 different motifs against the traditional Iznik tile patterns shown in Figure 1, it was concluded that both AI tools were knowledgeable about the penç, palmet, tulip, carnation, rose, spring branch, and tree of life motifs, producing comparatively satisfactory results with these prompts.

Comparing the two AI tools, we found out that DALL-E 2 produced results more consistent with the colors and patterns of Iznik tiles compared to Copilot. Copilot’s images featured more shading and depth, while DALL-E 2 produced more stylized visuals.

Consequently, it was decided to use the penç, palmet, tulip, carnation, rose, spring branch, and tree of life motifs within the scope of this study.

To determine the prompts to be entered into the AI tools, a step-by-step prompt creation method was developed. This method allows for testing different features in the AIs by constructing prompts from a simple grammatical structure composed of relevant adjectives and nouns to more complex expressions composed of combinations of descriptive phrases. Desired features were added incrementally as adjectives, and the sentence structure was tested step by step with the AI tools, monitoring their visual outputs. The prompts, prepared as noun phrases with the subject “An Iznik tile”, had adjectives added to specify the desired tile design in the order indicated in

Table 1. Adjectives of Iznik tiles.

IZNIK TILES ADJECTIVES
Order of Addition to a Sentence	1	2	3	4	5
Adjective Type	color	pattern geometry/composition	tile shape	motif	board shape
Adjectives	White	symmetrical	square	rose	circle
	Cobalt	asymmetrical	rectangle	carnation	square
	turquoise	continuing tiles/pattern (ulama)	hexagonal	tulip	rectangle
	Green	panel		spring branches	polygon
	coral red/tomato red			tree of life
	other colors			rumi
				haliç işi
				palmette (hatayi)/rosettes (stylized flower)
				chintemani (çintemani)

. The sequence of sentence elements was written to form a meaningful whole. Subsequently, adjectives describing the spaces in which the desired Iznik tiles were to be designed were added to the sentence according to the order specified in

Table 2. Adjectives of space.

SPACE ADJECTIVES
Order of Addition to a Sentence	1	2	3	4	5	6
Adjective Type	placement in space	function	open/closed space	building type	style	according to the mood
Adjectives	wall	metro/office	open space	private interior	modern architecture	sad
	floor	museum/exhibition/pavilion	semi-open space	private open space	contemporary	exciting
	ceiling	library	closed space	semi-private	Nordic	historically referenced
	carrier system	bank		public	country	dynamic
	position according to architectural elements	education structure		public interior	Victorian
	fixed architectural elements (fireplace, counter, niche, …)	train station			minimalist
	upper parts of the windows on one wall	cafe/restaurant/bar
	giving a principle for its traditional use in relation to its place in buildings	hotel/bath house/spa
	human eye level on one wall	sports buildings
	tile can be used on the surface where it is desired to create a focus in the space (in general, tiles are used in the most prominent/important architectural elements)	administrative buildings
	tile compositions can be positioned according to windows and used in relation to rows of windows

. Starting with the first added adjective, the sentence structure was tested at each stage with the AI tools, and the visual outputs were monitored step by step. The results were compared both within the generating AI tool and between different AI tools, allowing for thorough observations.

In line with the described method, two prompt series were created. Creating two different prompt series diversified our sample and increased the number of examples, allowing us to obtain more robust analyses. The first prompt is as follows: “A circle-formed Iznik tile panel with a tulip motif consisting of rectangular modules with white and cobalt colors, located on the upper parts of the windows on one wall of the semi-open area of the public interior of the train station in the city, minimalist style, dynamic”. The second prompt created is as follows: “A rectangle-formed Iznik tile pattern with a rose motif consisting of hexagonal modules with turquoise and coral red colors, located on the human eye level on one wall of the open space of the private interior of the office in the city, contemporary style, historically-referenced”. These two prompts were tested using four different AI tools: Copilot, DALL-E 2, DALL-E 3, and V6 Midjourney according to the specified method, and the results were analyzed. The results of the tests can be seen in the supplementary material (Supplementary S1, Supplementary S2).

The implementation phase of the prompts proceeded as follows with the initial prompt. On the selected artificial intelligence tool, the prompts were tested sequentially, starting with “An Iznik tile panel with white and cobalt colors” followed by “An Iznik tile panel consisting of rectangular modules with white and cobalt colors “and then “An Iznik tile panel with a tulip motif consisting of rectangular modules with white and cobalt colors”, “A circle formed Iznik tile panel with tulip motif consisting of rectangular modules with white and cobalt colors”, “A circle formed Iznik tile panel with tulip motif consisting of rectangular modules with white and cobalt colors on one wall”, “A circle formed Iznik tile panel with tulip motif consisting of rectangular modules with white and cobalt colors on one wall of a train station”, “A circle formed Iznik tile panel with tulip motif consisting of rectangular modules with white and cobalt colors on one wall of the semi-open area in the train station”, “A circle formed Iznik tile panel with tulip motif consisting of rectangular modules with white and cobalt colors on one wall of the semi-open area of the public interior of the train station”, “A circle formed Iznik tile panel with tulip motif consisting of rectangular modules with white and cobalt colors on one wall of the semi-open area of the public interior of the train station in the city”, “A circle formed Iznik tile panel with tulip motif consisting of rectangular modules with white and cobalt colors, located on the upper parts of the windows on one wall of the semi-open area of the public interior of the train station in the city”, “A circle formed Iznik tile panel with tulip motif consisting of rectangular modules with white and cobalt colors, located on the upper parts of the windows on one wall of the semi-open area of public-interior of the train station in the city, minimalist style”, and, finally, “A circle-formed Iznik tile panel with a tulip motif consisting of rectangular modules with white and cobalt colors, located on the upper parts of the windows on one wall of the semi-open area of the public interior of the train station in the city, minimalist style, dynamic”. The same procedure was repeated for the other prompt, with adjectives added to the sentence in the order specified in the method.

4.5. Evaluation

The images generated by AI in response to the experimental prompts explained above were evaluated by taking into account the design tradition of Iznik tiles and its qualified contemporary interpretations. A comparative analysis was conducted among the AI-generated images. The colors in the AI-generated images were compared to the traditional Iznik tile colors documented with RAL color codes in the study by Çorakbaş et al. (2024) [13]. The patterns produced by the AI software were also examined in terms of the similarities to and differences from the traditional Iznik tile patterns, examples of which are provided in Figure 1. Additionally, how Iznik tiles were represented within architectural spaces in the AI-generated images was compared to traditional and contemporary uses of Iznik tiles in space (Figure 3 and Figure 4). In Figure 3a,b, and Figure 4b, we see examples of the traditional and contemporary use of Iznik tiles in circular forms, as requested in the first prompt series. Similarly, in Figure 3b,c, there are examples of the use of traditional Iznik tiles on windows, which was requested in the first prompt. In Figure 4c, a tile panel produced with traditional references from a train station, as requested in the first prompt series, is visible. In Figure 4a, on the other hand, a contemporary panel with historical references in an open space, as requested in the second prompt, is displayed. We evaluated the final visual AI outputs by comparing them with real-life cultural heritage examples with similar characteristics.

In order to be able to comprehend whether AI-generated Iznik tile images reflect any concern on the relative dimensions of tiles and spaces, the dimensions of the panels in the obtained images, as well as the dimensions of the unit tiles forming these panels, were approximately calculated. As a method, the images were imported into 2024 AutoCAD software, where the grid structures of the tiles composing the panels were drawn following the visible joints on the ceramic panels as reference points. As a method, the images were imported into 2024 AutoCAD software, where the grid structures of the tiles composing the panels were drawn following the visible joints on the ceramic panels as reference points. At this stage, we observed that, in contrast to what we had expected, the joints of the panels were distinct and repeated in only a few of the AI-generated images.

The dimensions of the tiles in the AI-generated images were calculated approximately by attributing heights to architectural elements or the existing furniture. For instance, in the image shown in Figure 5b, the height of the table was estimated to be 75 cm and, using the vanishing lines in the one-point perspective drawing, this measurement was transferred to the level of the panel. The ceramic panel was then scaled in the CAD software using the “scale-reference” command. In Figure 5c, for scaling and approximately dimensioning the ceramic panel in the AI-generated image, a similar method was applied, and, as a beginning reference, the windowsill level was estimated to be 90 cm. Similarly, in the image shown in Figure 5d, the height of the railings on the right and left sides was estimated to be 90 cm, and this measurement was used as a reference to scale the ceramic panel. In Figure 5a, the height of the wall on which the panel is located was estimated to be 450 cm, and the ceramic panel was scaled according to this reference. Finally, based on the scaled drawings, the approximate dimensions of the ceramic panels and the unit tiles composing them were measured using the CAD software (Figure 6). This analysis was applied to the images generated by V6 Midjourney, DALL-E 3, and Copilot AI tools (Figure 5). However, DALL-E 2 was excluded from this analysis due to the lack of reference architectural elements in the images that it produced.

5. Results and Discussion

As explained in the methodology section, four different AI tools were used during this research period. It should be noted that one of these tools (DALL-E 2) has been discontinued due to an update. First, the basic similarities and differences between these four AI tools were discussed. Comparing the AI-generated images, we observed that the pattern and space images produced in Microsoft Copilot and DALL-E 3 were similar. The fact that Copilot uses the DALL-E 3 infrastructure proved to be the main reason for this situation. These two AI tools produced similar images in terms of color, pattern, and composition qualities, and no major differences were observed in the generated images.

DALL-E 2 and Midjourney, on the other hand, have made a wide variety of designs both in the alternatives produced for a specific prompt and in general throughout the prompt series. The inspiring aspect of these design proposals is higher and offers numerous alternatives to the user. Although DALL-E 2 and Midjourney are similar in terms of pattern generation and variety, they differ in terms of representing the space or interpreting spatial textual terms. It is a result of our study that AI tools generate visuals in different types as stated in [33,37]. The prompt expressions concerning architectural space did not result in images representing spaces in the DALL-E 2 application, and only tile pattern images were generated. Midjourney, on the other hand, has managed to generate various architectural spaces according to the prompts entered (Figure 7). Our results reveal that DALL-E fails to provide recommendations that integrate with the space, as observed by Moral-Andres et al., whereas Midjourney demonstrates greater efficiency in recommending the positioning of tile panels within the space.

Concerning the interpretation of traditional Iznik tiles in AI-generated images, DALL-E 2 and Midjourney tools are more successful in expressing vivid and bright colors, which is one of the unique qualities of traditional Iznik tiles. Especially for the red color, the red tones used in DALL-E 2-generated images were close to the traditional Iznik tile coral red tones (Figure 8b). In contrast to the results of the study of [46], which discusses that AI software can successfully reconstruct colors and shapes in the case of ancient mosaics, our research shows that the AI-generated images’ depiction of Iznik tile colors does not match with the traditional color palette of Iznik tiles, which is one of the most prominent characteristics of this unique tradition. As an exception, the images generated by DALL-E demonstrate more ability to represent the original color palette of Iznik tiles (Figure 8).

The findings of our study show that the different software exhibits different levels of ability in terms of fulfilling the design necessities of Iznik tiles, like pattern variety, creativity, and artistic visualization [37].

In terms of the relationship between color and space, we found that, despite no specific color being defined for the elements of the architectural space except the tile surface, the Midjourney tool used colors on various surfaces or furniture within the space that harmonized with the tile design, proposing a consistent color palette for space, thus better integrating the tile surface with the modern interior (Figure 9). This outcome corroborates the earlier finding that Midjourney possesses superior capabilities in spatial integration. In DALL-E 3, another tool that is successful in producing space visuals, as well as in some of Midjourney’s productions, gray and white colors were predominantly used in the spaces (Figure 10). The reason for this appears to stem from expressions providing public spaces such as workplace, train station (function), and minimalist, contemporary (style) in the prompt sequence. In the images generated by Midjourney, it was observed that natural and artificial light ambiances were depicted in the images, reinforcing spatial integration. Moreover, color and the reflective properties of the tile were emphasized in certain recommendations (Figure 10b). Consequently, it can be asserted that Midjourney possesses the potential to suggest images tailored to the design choices of architects.

In this context, these applications help to visualize and develop the image in the designer’s mind and convey the thoughts and conclusions of the architect to users and employers. Based on both the AI experimentations and the design office experiences of the authors, AI-supported programs can enable the design process to be inspired by unexpected responses to design problems and to be developed faster and more efficiently by visualizing the possible outputs. Additionally, AI also provides an advantage in managing the design process more efficiently by visualizing the different form and pattern alternatives.

Concerning the interpretation of the terms related to the motifs of the traditional Iznik tiles, the Copilot and DALL-E 3 tools visualized tulip and rose motifs in a way that was directly detected (Figure 11 and Figure 12). Both tools maintained their understanding of motif design throughout the prompt sequence. The DALL-E 2 and Midjourney tools were able to produce a variety of motif visuals. In particular, DALL-E 2 has created original motif images. A significant discovery about the motifs is that the tulip motif has yielded more stylized and varied compositions than the rose motif. The difference in the reproduction of these motifs can be explained by the limitations of AI technologies in reproducing traditional arts with unique qualities varying according to cultural codes as stated by Sukkar et al. (2024) [48]. The results indicate that the identification capacity of AI should be improved regarding cultural components. The identification of the tulip by AI tools in a vertical segment may account for this occurrence (Figure 12). This situation allowed for one-way compositions as well as the production of central and free compositions. The rose, on the other hand, is defined through its horizontal section in almost all the images produced. Thus, compositions formed by the repetition of the same motif are predominant.

Concerning the tile shapes generated in the AI images, quadrilateral and hexagonal tile forms were tried in prompt sequences. In general, all AI tools created images responding properly to the tile shapes defined in the prompts. However, rarely, there are also image examples where the tile shape is undefined, and different polygons are used instead of hexagons in the hexagonal prompt sequence (Figure 13). A remarkable feature of the images produced by Midjourney is the rendering of weathered or deformed images on some of the tiles (Figure 14). This image can be expressed as the value of aging. Although no prompt has been entered to create this image, it can be said that Midjourney produced the Iznik tile with reference to existing images of traditional uses or defined the Iznik tile with its antiquity value.

In terms of pattern geometry and composition, in the prompts created, the expression of the panel was used for defining the tile compositions on panels. Therefore, it was envisaged that it would cover a limited area on an architectural surface within space. According to the results obtained, AI tools except DALL-E 2 achieved this. DALL-E 2 could not create data about the space; rather, the images that it generated included infinite patterns (Figure 15b). Generally, symmetrical order was employed in the compositions produced by AI. Additionally, DALL-E 2 and Midjourney produced asymmetrical compositions in several proposals. In the second prompt, which required hexagonal tiles with tulip motifs, repetitive (pattern-based) compositions formed the majority.

Examining the grid structure and tile dimensions of the ceramic tiles produced by AI tools, we observed that the grids were often unclear and not easily perceived in most of the AI-generated images. However, analyses of images where the grid structure was identifiable revealed that AI tools generally produced results similar to the existing and most-used tile shapes in real life, that is, squarish and rectangular tiles. In cases involving longer and more detailed prompts, the grids became more aligned with the intended outcomes. Nevertheless, in the second set of prompts requesting hexagonal tiles, while Copilot and DALL-E 3 generated octagonal and complex geometric shapes, Midjourney and DALL-E 2 delivered results closer to the desired design.

The analysis of the tile dimensions shows that the hexagonal tiles produced by Midjourney had an edge length of approximately 11.5 cm (Figure 16c). In a Midjourney-produced image, the panel consisted of 13 cm square tiles and 13 × 9.4 cm rectangular tiles (Figure 16b), whereas a DALL-E 3-created image included 21.3 × 12.3 cm rectangular tiles (Figure 16c). On the other hand, an image generated by Copilot depicted 16.6 cm square tiles (Figure 16d). Although these dimensions are close to the traditional Iznik tile sizes, they were found to be slightly smaller. In Figure 17a, Copilot generated a circular panel with a diameter of 285 cm, while, in Figure 17d, DALL-E 3 produced a circular panel with a diameter of 196.9 cm. Additionally, in Figure 17d, Midjourney created a rectangular panel with dimensions of 171 × 346 cm, while, in Figure 17c, it produced an arch-shaped panel with dimensions of 240.1 × 270.7 cm. The panel dimensions generated by AI are generally larger than traditional panel sizes. It is remarkable that, in the AI-generated images, unlike real-life examples, there is no clear relationship between panels and the tiles composing them in terms of shape and size. Another result of our analysis of the dimensions and shapes of the panels in the AI-generated images is that panel sizes and scales and their relationships with the other architectural elements in the space vary to a high extent depending on the architectural context in which they are used, as is the case in real-life traditional examples.

To conclude, we claim that AI tools may offer design suggestions and inspirations in architectural and element scale and can play an auxiliary role in accelerating and enriching the design process, while they may be misleading in the reinterpretation of traditional arts and crafts with their correct authentic references to material and space. Therefore, architects should approach the reinterpretations of traditional arts and crafts by AI software with a critical perspective.

Concerning the placement of tile compositions in space, in all prompt series, we requested text-to-image AI platforms to generate images with tile compositions on a wall surface as, in the field of application, the tile design requests usually involve wall surfaces. All the AI tools were able to produce design proposals for a wall. Regarding the place of the tiles on the wall, the top of the windows was defined in the first prompt series while the human eye level was defined in the second prompt series. For the human eye level, it is observed that, in the generated images, tile compositions started from the floor and reached the ceiling of the space (Figure 18). This issue has led us to believe that AI technologies are incapable of understanding the concept of “human eye level” inside a given environment. Within the series of prompts that required photographs of tiled surfaces at the window level, there were AI-generated ideas that fulfilled this requirement, as well as images that did not adhere to the prompt’s specifications. Again, in this prompt series, all images generated by Midjourney managed to produce a variety of proposals concerning the place of tile compositions in space. It was remarkable, in one of the images that was produced by Midjourney, that similar patterns on a carpet surface on the floor and a tile panel surface on a wall were used (Figure 9), even though the task did not specifically require it. We believe that this is due to Midjourney’s ability to generate multiple and varied demands for textual expressions, resulting in more pleasing and cohesive visuals. Therefore, Midjourney’s utilization of the requested components is fractal in nature, rather than following a Cartesian approach. This approach emphasizes Midjourney as a supportive program in the design process for interpreting repeating geometric compositions in relation to architectural space. We also noted that, in DALL-E 3, in addition to the depiction of tile surfaces on the walls, as was required by the prompts, tile compositions were also used on floors and ceilings.

In terms of the responses of AI to the defined building types and interior or exterior spaces, AI tools’ responses to expressions such as private interior and semi-open area were tested through prompts. We evaluated that AI tools detect these expressions and, when we examined the images that they produced in detail, we observed that they contain details responding to the spatial terms. We also saw that the more the space is defined in detail, the more the result satisfies the designers’ expectations. Our results support the argument that it is important to consider prompt length as stated by Chang et al. [39] and to express it in detail. It may also be important to elaborate prompt expressions in terms of the function of the space and the place of the tile compositions in the space.

Within the context of this research, the AI’s responses to the architectural style definitions used in the prompts were focused on showcasing the potential of incorporating traditional tile designs with innovative approaches in contemporary architecture. As a result, contemporary and minimalist expressions were given preference. According to the image results of the prompts that include style expression, we discovered that Midjourney, which is among the most widely used AI tools, could generate a variety of relevant images that had the potential to inspire the designer in multiple ways. Our research supports the existing findings of research [33,37] that AI tools perceive visual outcomes through unique manifestations of creativity and artistic visualization. Additionally, our findings highlight the variations in performance in design assistance among these tools. For instance, although DALL-E 2 could quickly generate images in response to the prompts consisting of design inquiries, the details in the images produced by DALL-E 2 were not clear. Therefore, the contribution of the DALL-E 2-generated images remained at an inspirational level. On the other hand, DALL-E 3, Copilot, and Midjourney produced better results than DALL-E 2 in creating images depicting architectural spaces. The results indicate that the choice of certain software or a combination of software can expedite and enhance the design process following architectural decisions and objectives.

Our results also indicate that various AI tools display potential for future advancements in the restoration and conservation of historical buildings and the contemporary interpretation of cultural and local elements in the design of modern interiors, as reported in the literature [42,44,46,47]. Within this framework, the designer can generate efficient ceramic and space design proposals more quickly by selecting one or several AI tools that are most appropriate in terms of color, pattern, style, and composition. It is important to highlight that providing more elaborate language and prompts for decisions regarding design results in more successful outcomes.

6. Conclusions

Iznik tiles are commonly present in both public and private buildings, at various architectural and urban levels, generally elaborating interiors, facades, entrances, and urban facades. Ceramics and Iznik tiles are utilized in the ornamentation of new buildings, i.e., congress halls, banks, and transportation structures, as well as in the restoration of existing historical buildings. They can be incorporated into architectural and urban designs at various scales. A unique characteristic of Iznik tiles is their seamless incorporation into the architectural environment, a quality that has become increasingly rare in modern designs of late. This project investigates the potential of utilizing artificial intelligence to design Iznik tiles in a manner that aligns with the overall spatial aesthetics, departing from the conventional approach of incorporating disconnected panels with repetitive traditional patterns. Additionally, this study discusses how AI can contribute to contemporary designs and traditional tile designs. The following conclusions are reached within the scope of this study:

• AI can assist ceramic tile designers throughout the initial phases of the design process by producing creative and diverse thoughts.
• AI can assess the traditional patterns employed in Iznik tiles. Artificial intelligence can examine and interpret these artistic styles and structures and create novel and creative techniques that connect this traditional art with the contemporary world.
• Artificial intelligence can enhance and streamline the tile manufacturing process. While AI might be advantageous in economically managing the production process, architects must prioritize the suitability of visual aspects of materials, ornamentation, and architectural choices.
• Upon examining the visuals obtained from AI programs, it can be observed that some tools excel in diversity and creativity while others are better at artistic visualization; nevertheless, these tools are not successful in creating visuals that incorporate intangible values such as cultural codes. Furthermore, the fast reaction and user-friendly nature of many AI software programs allow for integration, enabling designers or employers to use the programs together or select a program based on architectural requirements or user preferences.
• While AI models do not neglect fundamental principles such as aesthetics, functionality, and sustainability in integrating tiles with space, they are not successful in analyzing expressions related to user psychology. This can be seen as a limitation, but it can also be considered as a positive situation for the justification of the necessity of the designer’s role and the need for interpretation in the design process.
• AI can provide suggestions on how Iznik tiles can be used in the design of spaces. For example, it can analyze which patterns and colors are more suitable for specific walls or areas. This is a contribution that can lead to a faster and more successful design of different spatial scales within short deadlines. In addition, it can also contribute to the production of samples of Iznik tiles, which have a unique production process, before the implementation. This is also useful for identifying a possible problem in the application process in advance.
• Midjourney may be used as one of the artificial intelligence tools for generating images related to tile designs; may offer different compositions; and may be inspiring for the reinterpretation of traditional colors in tile designs. The fact that it can produce integrated designs that can respond to the expressions about the space shows that it can be used to strengthen the tile–space relationship, and it is, in this respect, more successful than other AI tools in terms of understanding the patterns, performance in spatial representation, accurate representation of colors, and reflection of the contemporary architectural approach (

  Table 3. Comparison of AI tools.

  Feature	DALL-E 2	Copilot (Microsoft Image Creator)	DALL-E 3	Midjourney
  Launch date	April 2022	October 2023	September 2023	July 2022
  Image resolution	1024 × 1024	1024 × 1024	1024 × 1024/1792 × 1024	2048 × 1280
  The performance in understanding the patterns of Iznik tiles	good	moderate	moderate	good
  The performance in representing Iznik tiles within the space	poor	moderate	moderate	good
  Accurate representation of Iznik tile colors	good	moderate	good	good
  Perception and reflection of contemporary architectural design concepts	poor	moderate	moderate	good

  ).
• Although DALL-E 2 could not visualize the expressions of the space and did not generate images that show compositions in space although the prompts asked to do so, it was quite inspiring in terms of the images that it generated regarding the tile pattern and color design, which related traditional and modern or contemporary styles (Table 3). Consequently, it appeared to possess significant inspirational potential for the designer. It may be asserted that DALL-E 3, because of its enhanced capabilities, serves as a superior assistant in the design process compared to DALL-E 2.
• The DALL-E 3 and Copilot tools produced images that were very similar in terms of color palette and patterns, whereas DALL-E 3 was better at visualizing spatial data. Upon assessment, it was determined that both software algorithms produced images devoid of the characteristic attributes of traditional Iznik tiles, including colors, patterns, and representations.

In light of the aforementioned contrasts, Midjourney seems to be the most contributing program in terms of design composition for space designs in the interior, façade, or other surfaces, or for developing experimental alternatives in historical surroundings. However, DALL-E 2, with better performance in experimenting with different styles in terms of pattern and color, has the potential to contribute to the designer making clearer decisions by providing a successful visualization for the designer. Simultaneously, it might aid the designer in effectively conveying the design concept to the user or employer. However, DALL-E 3 and Copilot do not offer sufficient favorable feedback during the design phase. Nonetheless, they do have the benefit of aiding in the modeling of the tile composition in the proposed space.

Based on the fact that AI can support the design process, producing several options rapidly is advantageous for designing buildings with diverse purposes. Moreover, the ability to conceptualize and visualize alternatives within the designer–user–employer dynamic is highly helpful for both the design and implementation phases. Through the visualization of many traditional patterns at the architectural scale, AI may improve the decision-making process for tile designs that more effectively integrate with the architectural space. When AI is further developed in the identification of cultural codes, it may improve its ability to provide tile or architectural design recommendations that are aligned with the socio-cultural context. Identifying software that aligns more effectively with objectives is also a significant area of study that will enhance both theoretical and practical domains.

In terms of architectural and urban conservation, on the other hand, the proficiency and advancing efficacy of AI in the recognition and classification of tangible cultural heritage elements can enhance how AI can be useful for the conservation of tile heritage. Below, some of the potential uses of AI in cultural heritage conservation are listed:

• AI tools’ success in recognizing and classifying cultural heritage elements can be supportive in developing proper conservation decisions for cultural heritage at architectural and urban scales [69].
• Another contribution of AI in the conservation of cultural heritage is its success in the recognition of visual traces of material deterioration, which can help in developing surface and tile conservation and repair.
• Enhanced analysis and classification of cultural heritage items through AI can facilitate the creation of more effective heritage interpretation and presentation strategies, especially concerning museum collections or cultural heritage at archives or archaeological sites.
• By correctly visualizing the reinterpretation of traditional arts and crafts in contemporary spaces, AI tools can play a crucial role in the distribution and awareness of—especially aesthetic—values of cultural heritage, enhancing the need for plural interpretations and presentations of heritage assets.
• Possibilities for the participation of communities in the architectural or urban conservation processes are provided [70].

On the other hand, it should be noted that the conservation of architectural and urban heritage is a highly complex process [71], requiring thorough studies on deciphering the relationships of cultural heritage with nature, history, society, and politics [72]; thus, it should be managed by well-equipped experts. In the whole process of conservation, which requires analyses in various scales and qualities, integrated approaches combining communities, architectural and urban spaces, tangible and intangible cultural qualities, and natural elements, the contribution of AI should be well defined and strictly controlled and supervised by experts who are educated in the integrated use of architectural and urban conservation methods, as well as in AI technologies.

In general terms, the rapid and user-friendly interfaces of AI software enable architects to incorporate this technology into the design process. AI’s ability to design, visualize, and provide multiple options can be regarded as an extremely useful contribution also to both the designer–user–employer relationship and economic limitations. Many software programs’ fast reactions and simple interface make integration easy, allowing designers or employers to use them together or choose one depending on their architectural demands or user preferences. In this context, AI has the capacity to be a tool that enhances the efficiency and speed of the design and implementation process for Iznik tiles. It aligns with the designer’s decisions and user demands while effectively managing economic resources. This is particularly important considering the expensive production process of Iznik tile. Regarding the design of Iznik tiles in architecture, which are preferred in public as well as private buildings, official authorities act as both the employer and the user, complicating the persuasive and decision-making stages of the design process. The ability of AI software to produce varied possibilities rapidly and contextually can improve the design and implementation phases by assisting in architectural decisions.

In conclusion, this research has revealed that the current state of text-to-image AI tools needs to be improved in terms of their understanding of specific terms related to traditional design forms and patterns. This improvement is necessary for these tools to effectively assist in integrating traditional and contemporary designs within the field of architectural and interior spatial production. This research also demonstrates that the designer, who plays a crucial role in determining human needs and the AI-supported design process, should possess extensive knowledge of both traditional and contemporary design elements, in terms of both language and visuals, in order to make informed decisions when selecting AI-generated alternatives. Hence, although AI can save time and enhance efficiency in the design process, it is crucial to acknowledge that a designer’s limited expertise may hinder their capacity to assess the appropriateness of AI-generated images for effective design. The proficient application of AI by architects in the design process, as well as the integration of AI technology, is a rapidly evolving and emerging field. Future research opportunities encompass the enhancement of AI algorithms to improve program outputs within cultural and historical contexts, facilitating the generation of design concepts for architectural and urban conservation projects and the repair of deteriorated or damaged architectural components.

Another essential area for future research is to enhance the comprehension of specialized terminology related to architectural expressions and ceramic and tile descriptors, as well as psychological mood adjectives. This could improve AI outputs that aid in the design process by including user preferences, spatial dynamics, and architectural components. The development of AI tools for examining psychological responses, including emotional thoughts and preferences for patterns and colors in Iznik tiles, would be a significant advancement in the fields of architecture and cultural heritage conservation. Additionally, some areas of interest involve customizing AI tools to learn from psychological responses, user feedback, and designer decisions, improving the linguistic understanding of AI-supported design processes in texts and prompts, and analyzing the economic benefits of implementing AI in terms of time, production, and efficiency. Future research may also focus on assessing AI software and making comparisons between AI tools for making aesthetic or functional recommendations. The integration of aesthetic components with fundamental design principles in spatial design can enable an assessment of the visual outputs of various software, hence aiding in the enhancement of training datasets in accordance with specific design principles. In addition to evaluating aesthetic elements, analyzing AI outputs within functional and architectural frameworks in relation to the designer’s intent will improve the evaluation of the software’s efficacy in practical applications and guide its development through results-oriented training. AI, which continuously evolves through varied datasets, can be trained on tangible and intangible values of architectural and urban heritage and designs from various geographical areas. Enhancing AI’s text-to-image infrastructure can contribute to its efficiency in conservation and restoration while improving the understanding of local cultural and architectural elements. In this context, the data entry on various cultural qualities of heritage sites, along with the compilation of AI-assisted design recommendations in these areas, would enhance the AI software’s performance and expedite its growth. Future studies may utilize AI across varying geographical regions to develop design ideas, thus enhancing its adaptability to changing contexts.