Integrating Generative AI in Architectural Education: A Comparative

Study of Traditional, Stock LLMs, and Custom Tools

Abdelrahman Aly, Abdelsamie Elazazy and Nada Sharaf

Informatics and Computer Science, German International University, Cairo, Egypt

Keywords:

Generative AI, Architectural Education, ArchAI, Design Automation, Creativity, Building Information

Modeling (BIM), Educational Technology, Architecture Learning Outcomes, AI-Assisted Workﬂows.

Abstract:

The rapid development of generative artiﬁcial intelligence (AI) is transforming architectural education by re-

shaping creativity, technical skills, and problem-solving approaches. This paper presents a comparative analy-

sis of traditional methods, general-purpose AI tools like ChatGPT and Midjourney, and a custom-built Archi-

tecture AI Tool (ArchAI) tailored to the needs of architectural education. The study highlights the strengths and

limitations of each approach, focusing on their impact on creativity, efﬁciency, and educational outcomes. The

ﬁndings reveal that while general-purpose AI tools enhance accessibility and ideation, their domain-speciﬁc

applications are limited. In contrast, custom AI solutions, integrated with architectural principles and tailored

datasets, offer signiﬁcant advantages by automating design tasks, providing real-time feedback, and fostering

innovative learning experiences. This work underscores the need for a balanced integration of generative AI

to optimize learning outcomes and prepare students for professional practice.

1 INTRODUCTION

The fast-paced advancement of artiﬁcial intelligence

(AI) has introduced groundbreaking tools that are

transforming various ﬁelds, with generative AI stand-

ing out as a key technology in creative industries.

Using sophisticated computational methods such as

large language models (LLMs), generative adversar-

ial networks (GANs), and diffusion models, genera-

tive AI facilitates the automated creation of diverse

content, including text, images, 3D models, and ar-

chitectural designs. This technological leap carries

signiﬁcant implications for architecture a ﬁeld tradi-

tionally dependent on human creativity, iterative pro-

cesses, and extensive technical expertise. In architec-

tural education, generative AI is not only challeng-

ing traditional learning frameworks but also reshaping

how creativity, problem-solving, and technical skills

are nurtured (Li et al., 2024a).

At its most accessible level, generative AI is rep-

resented in popular tools like ChatGPT or image gen-

erators such as Midjourney, designed for general-

purpose use. These tools are user-friendly and ver-

satile, making them valuable for early-stage ideation,

content creation, and visual exploration in architec-

tural projects (Ploennigs and Berger, 2023). For in-

stance, ChatGPT has been used to craft narratives

around architectural concepts or reﬁne project de-

scriptions, while platforms like DALL-E and Mid-

journey allow students to quickly visualize abstract

ideas, streamlining the early design phase. By low-

ering the barrier to AI-driven creativity, these tools

enable even beginners to experiment with sophisti-

cated design ideas. However, the general-purpose

nature of these tools often falls short in addressing

the specialized demands of architecture. They may

lack the precision, contextual awareness, or depth

needed to tackle complex architectural challenges.

Recent works have shown how targeted applications

of generative AI can bridge this gap, as exempli-

ﬁed by its application in the building industry to im-

prove workﬂows and enhance code compliance (Wan

et al., 2024). In contrast, custom-built generative

AI tools offer tailored solutions that align with the

speciﬁc needs of architecture students and educators.

These tools, designed with domain-speciﬁc datasets,

architectural principles, and educational objectives in

mind, can integrate seamlessly with Building Infor-

mation Modeling (BIM) systems or provide design

feedback consistent with professional standards (Li

et al., 2024b).

Users generally favor tools that are user-friendly

and customizable for visualizing and presenting di-

verse data types (Roshdy et al., 2018; Sharaf et al.,

Aly, A., Elazazy, A. and Sharaf, N.

Integrating Generative AI in Architectural Education: A Comparative Study of Traditional, Stock LLMs, and Custom Tools.

DOI: 10.5220/0013378000003912

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2025) - Volume 1: GRAPP, HUCAPP

and IVAPP, pages 415-420

ISBN: 978-989-758-728-3; ISSN: 2184-4321

415

2014). This preference extends to architectural plans

as well. Additionally, tools that enable new ways of

interacting with these plans can be particularly ben-

eﬁcial. The potential of generative AI in architec-

tural education goes far beyond automation. Research

suggests that carefully integrating these tools can en-

hance creativity and improve time management. For

example, studies have shown that architecture stu-

dents often use generative AI during the concept de-

velopment stage to visualize ideas and streamline

workﬂows, leading to clearer concepts and greater

efﬁciency (Kee et al., 2024). Moreover, AI-driven

tools like ArchiGuesser have demonstrated the abil-

ity to make learning architectural history more in-

teractive and engaging through gamiﬁed experiences

(Ploennigs and Berger, 2023).Generative AI’s capa-

bilities extend further, as shown in its ability to gener-

ate multimodal outputs, such as text-to-image models,

which have been explored for educational and design

ideation purposes (Koh et al., 2023).

Despite these advantages, questions persist about

the effectiveness and ethical implications of using

generative AI in education. Comparisons between

traditional methods and AI-supported workﬂows re-

veal striking differences in how creativity, human

agency, and educational outcomes are approached.

Traditional methods, rooted in manual drafting and

iterative instructor feedback, often emphasize foun-

dational skills and critical thinking. On the other

hand, generative AI offers faster ideation and en-

hanced visualization, raising concerns about potential

over-reliance on automation and the erosion of essen-

tial skills. Additionally, within the realm of generative

AI, there is a great contrast between general-purpose

tools like ChatGPT and specialized AI systems built

for architectural applications. While the former prior-

itizes accessibility and ease of use, the latter delivers

targeted solutions, supporting tasks such as sketch-

to-model transformation, generative ﬂoor plans, and

context-aware feedback tailored to architectural edu-

cation (Li et al., 2024b).

This paper focuses on the contrasts between tradi-

tional methods, general-purpose generative AI tools,

and custom AI solutions. By examining the impact of

these approaches on architectural education, it aims

to highlight the most effective ways generative AI can

enhance learning outcomes.

2 METHODOLOGY

This study employs a qualitative framework to eval-

uate the role of generative artiﬁcial intelligence

(GenAI) in architectural education. The methodology

involves a comparative analysis of traditional non-AI

approaches, generative AI tools, and the custom-built

Architecture AI Tool (ArchAI), emphasizing their

respective contributions to creativity, learning efﬁ-

ciency, and design workﬂows. Additionally, the study

contrasts stock generative AI solutions, such as Chat-

GPT and Claude AI, with the custom tool to identify

domain-speciﬁc advantages. Images from the Archi-

tecture AI Tool (ArchAI) and its outputs will be inte-

grated for demonstrative purposes.

2.1 Comparative Framework:

Generative AI vs. Non-Generative

The study examines the differences between gener-

ative AI-enabled workﬂows and traditional non-AI-

supported methods. Generative AI offers signiﬁcant

advantages in automating repetitive tasks, providing

real-time feedback, and enhancing visualization ca-

pabilities. These aspects are compared against non-

generative approaches, which emphasize manual ef-

fort and critical thinking.

Firstly, in terms of Automated Design Assistance,

generative AI tools, such as the Architecture AI Tool

(ArchAI), automate blueprint analysis, layout sugges-

tions, and design visualization, allowing students to

focus on conceptual development. Non-generative

methods rely on manual iterations and instructor-led

feedback, which, while foundational, are less efﬁcient

for rapid prototyping. By manually adjusting layouts,

students spend more time learning technical processes

but less time innovating and exploring creative possi-

bilities.

Secondly, regarding Feedback and Iteration, the

Architecture AI Tool (ArchAI) provides actionable,

data-driven feedback linked to uploaded blueprints,

allowing immediate reﬁnement. This is achieved

by analyzing spatial layouts, identifying inefﬁcien-

cies, and suggesting practical improvements based

on its integrated architectural knowledge base. Tra-

ditional methods depend on instructor evaluations,

which, while valuable for mentorship and guidance,

are constrained by the availability of instructors and

often lack scalability for larger cohorts. The reliance

on periodic feedback slows the iteration process and

reduces the agility of student workﬂows.

Thirdly, in the aspect of Visual Outputs, gener-

ative AI tools generate updated images and alterna-

tive layouts from user inputs, enhancing creativity

through visual exploration. This visual immediacy

enables students to evaluate multiple design iterations

in real-time. Non-generative methods, though rein-

forcing manual drafting skills, require extensive effort

GRAPP 2025 - 20th International Conference on Computer Graphics Theory and Applications

416

to achieve comparable visual exploration. As a result,

students are often limited to fewer iterations within

constrained time frames.

By accelerating the design process and enhanc-

ing visual feedback, generative AI fosters a more dy-

namic learning environment while traditional meth-

ods build foundational skills critical for long-term ex-

pertise.

2.2 Comparative Analysis: Stock vs.

Custom Generative AI Tools

A critical component of this study is the comparison

between general-purpose generative AI tools (e.g.,

ChatGPT) and the custom Architecture AI Tool (Ar-

chAI). The analysis highlights how customization en-

hances functionality for domain-speciﬁc tasks in ar-

chitectural education.

Firstly, concerning Blueprint Interaction and

Processing, the Architecture AI Tool (ArchAI) pro-

cesses uploaded blueprints, allowing users to rename

rooms, edit layouts, and generate updated visual out-

puts. These capabilities encourage students to interact

directly with design materials, fostering a hands-on

learning experience. Stock generative AI tools, such

as ChatGPT, are limited to providing textual descrip-

tions or conceptual advice. They cannot directly inter-

act with visual inputs like blueprints, restricting their

applicability in architectural education.

Secondly, in terms of Domain-Speciﬁc Knowl-

edge Integration, the custom tool incorporates a cu-

rated architectural knowledge base, offering precise,

contextually relevant responses to queries. This en-

ables the tool to address speciﬁc design challenges,

such as spatial optimization or material selection,

with tailored recommendations. Generic AI tools rely

on broad datasets, which, while versatile, often lack

the depth needed for specialized applications. As a

result, their responses are less accurate and may re-

quire signiﬁcant interpretation by the user.

Thirdly, regarding Visual Design Suggestions,

the custom tool generates precise, visually accurate

layouts based on user inputs, ensuring alignment with

architectural principles. For instance, students can

request improved layouts and receive tailored visual

outputs that incorporate lighting, spacing, and func-

tional zoning considerations. Stock generative AI

tools, while capable of generating images (e.g., us-

ing DALL-E), lack architectural accuracy. These

tools often produce outputs that are visually appeal-

ing but fail to adhere to practical design constraints.

Moreover, they struggle with understanding domain-

speciﬁc queries.

Fourthly, considering Ease of Use and Adapt-

Figure 1: Tool output when asked to rename Lounge 11x12.

Figure 2: GPT output when asked to rename Lounge 11x12.

ability, the Architecture AI Tool (ArchAI) integrates

seamlessly into architectural workﬂows, providing

features such as targeted feedback and interactive

editing. Its user-centric design ensures accessibility

for both novice and advanced users. Stock AI tools,

Integrating Generative AI in Architectural Education: A Comparative Study of Traditional, Stock LLMs, and Custom Tools

417

Figure 3: Original Blueprint Query.

Figure 4: GPT output when asked to provide a better layout

based on the original blueprint.

Figure 5: Tool output when asked to provide a better layout

based on the original blueprint.

though ﬂexible and broadly applicable, require sig-

niﬁcant adaptation to align with the speciﬁc needs of

architecture students. This gap limits their usability

for domain-focused tasks.

Lastly, in terms of Actionable Feedback and

Metadata Analysis, the custom tool offers metadata

analysis of blueprints, providing insights into dimen-

sions, zoning, and other structural elements. This

level of detail supports advanced educational objec-

tives, such as teaching students to evaluate design

metrics effectively. Stock tools lack the ability to in-

terpret or analyze such data, making them unsuitable

for tasks requiring detailed architectural insight.

The customization of the Architecture AI Tool

(ArchAI) enables it to address the unique demands

of architectural education, surpassing the capabilities

of generic AI tools in delivering targeted, actionable

insights.

2.3 Educational Applications of the

Architecture AI Tool (ArchAI)

The Architecture AI Tool (ArchAI) offers transfor-

mative potential in architectural education by enhanc-

ing student engagement, creativity, and learning out-

comes. Key applications include:

Firstly, Interactive Blueprint Analysis: Students

can upload blueprints, receive targeted feedback, and

reﬁne their designs interactively. This fosters active

learning by allowing students to experiment with lay-

outs and assess the impact of changes in real time.

Secondly, Accelerated Design Processes: The tool

automates repetitive tasks such as renaming rooms,

generating layouts, and evaluating spatial efﬁciency.

By reducing the time spent on manual processes, stu-

dents can dedicate more effort to conceptual develop-

ment and innovation.

Thirdly, Knowledge Retrieval: Leveraging its

domain-speciﬁc knowledge base, the tool provides

precise answers to complex queries. For example, stu-

dents can inquire about optimal materials for sustain-

able designs or efﬁcient zoning strategies and receive

contextually relevant insights.

Fourthly, Creative Exploration: The tool’s ability

to generate new layouts and design suggestions en-

courages experimentation. Students are encouraged

to explore unconventional ideas and develop inno-

vative solutions without the constraints of traditional

workﬂows.

Lastly, Enhanced Visual Learning: By generating

visual outputs tied to speciﬁc design principles, the

tool supports visual learners in grasping complex ar-

chitectural concepts. This enhances comprehension

and retention, particularly for students who beneﬁt

from diagrammatic representations.

The integration of these applications into architec-

tural education equips students with practical skills

GRAPP 2025 - 20th International Conference on Computer Graphics Theory and Applications

418

and a deeper understanding of design principles,

preparing them for future professional challenges.

3 RESULTS

This section compares Generative AI tools with tra-

ditional non-generative methods in architectural de-

sign, highlighting the enhancements in speed, creativ-

ity, and effectiveness. We also assess the capabilities

of specialized tools like the Architecture AI Tool (Ar-

chAI) against standard generative AI tools. Insights

are derived from practical evaluations with architec-

ture students, focusing on usability and System Us-

ability Scale (SUS) scores, to demonstrate the trans-

formative impact of Generative AI in architecture.

3.1 Generative AI vs. Non-Generative

AI Approaches

Generative AI tools demonstrated a signiﬁcant ad-

vantage in accelerating the iterative design process.

These tools reduced the time required for revisions

compared to manual methods, enabling students to vi-

sualize and reﬁne multiple iterations in real time. In

contrast, traditional methods, while reinforcing core

design skills, constrained students to fewer iterations

due to the manual nature of the workﬂow and time

limitations.

The use of Generative AI workﬂows also en-

hanced creativity by providing real-time feedback and

actionable suggestions. This allowed students to ex-

plore innovative layouts and optimize designs with

greater conﬁdence. On the other hand, manual meth-

ods relied heavily on instructor feedback, which, al-

though valuable, was less scalable and often delayed

due to instructor availability.

Additionally, tools such as the Architecture AI

Tool (ArchAI) provided enhanced visual feedback

that enabled students to experiment with a wide range

of design variations. This immediacy encouraged ex-

ploration and iterative improvement. In comparison,

non-generative methods required external tools for vi-

sualization, which limited both the immediacy and

scope of design exploration.

3.2 Stock Generative AI vs. ArchAI

ArchAI, such as the Architecture AI Tool (ArchAI),

offered superior performance compared to stock gen-

erative AI tools, particularly in their ability to process

and edit blueprints. This functionality provided stu-

dents with a hands-on learning experience that stock

tools, limited to textual interactions, could not repli-

cate.

The custom tool also outperformed stock AI in

terms of domain-speciﬁc knowledge. By leverag-

ing its architectural knowledge base, the tool pro-

vided precise and actionable insights for speciﬁc de-

sign challenges. In contrast, stock tools often pro-

duced generalized outputs that required additional in-

terpretation to be useful in a practical context.

Finally, design accuracy and visualization were

signiﬁcantly improved with custom tools. These tools

generated contextually accurate layouts aligned with

architectural standards, offering practical applicabil-

ity. In comparison, stock tools, while visually ap-

pealing, frequently lacked adherence to design con-

straints, reducing their utility for architectural tasks.

3.3 Usability Testing and SUS Scores

Results

To conduct the results, we conducted usability and

SUS tests for the Architecture AI tool (ArchAI). The

evaluation involved 31 students, primarily students

from architecture, to evaluate ArchAI’s functionali-

ties. The users were ﬁrst asked about their experience

with similar tools, such as ChatGPT, Claude AI, and

PlanFinder. Among the 31 users, 11 had advanced

experience, 10 had intermediate experience, 9 were

beginners, and 1 had no prior experience with similar

tools. They tried out the tool and its complete features

and were then asked to give feedback on its usability.

Users were asked various questions, including

how easy it was to use some features of the tool,

which they gave an average score of 4.6 out of 5, as

well as how easy it was to use the tool. The processes

of renaming a room and their overall satisfaction with

the enhanced layout both received average scores of

4.4 out of 5.

None of the users encountered issues or errors

while using the tool, they all successfully renamed

and updated the ﬂoorplan without difﬁculties. Addi-

tionally, 96.8% of them said that the tool responded

with a helpful and enhanced ﬂoorplan. As for the

feedback of the uploaded ﬂoorplan, all the users

agreed that the feedback was clear and actionable.

When asked about the likelihood of using the tool

again, the majority of the users expressed high inter-

est, with many responding “Very Likely.”

Regarding the SUS scores, ArchAI achieved a

high average SUS score of 86.69, with a standard

deviation of 21.56, which corresponds to an “ex-

cellent” usability rating based on established bench-

marks. Users gave an average score of 4.68 for want-

ing to use the tool frequently, 4.74 for ease of use, and

Integrating Generative AI in Architectural Education: A Comparative Study of Traditional, Stock LLMs, and Custom Tools

419

4.87 for conﬁdence when using the system. Further-

more, the tool was rated highly for being well inte-

grated (4.77) and quick to learn (4.77). Meanwhile,

scores for negative aspects, such as perceived com-

plexity (1.81), inconsistency (1.87), and cumbersome

processes (1.74), were very low, indicating minimal

issues in these areas. These results reﬂect a very pos-

itive user experience overall, with high usability and

conﬁdence ratings, and low scores for complexity and

inconsistency. This metric shows the system’s effec-

tiveness and that it is user-friendly.

4 CONCLUSION

This study demonstrates that generative AI, particu-

larly custom-built solutions like the Architecture AI

Tool (ArchAI), can revolutionize architectural educa-

tion. These tools enhance creativity, streamline work-

ﬂows, and provide targeted feedback, allowing stu-

dents to focus on conceptual innovation. Compared to

traditional methods and generic AI solutions, the cus-

tom tool excels in addressing domain-speciﬁc chal-

lenges, offering superior interactivity, precision, and

contextual relevance.

ACKNOWLEDGEMENTS

We extend our gratitude to Dr. Maha ElGewely and

the faculty members of the Architecture Department

at GIU for their invaluable assistance with the focus

group and for offering various plans to evaluate the

environment. We also acknowledge the use of Ope-

nAI’s ChatGPT in the writing process of this paper.

REFERENCES

Kee, T., Kuys, B., and King, R. B. (2024). Generative arti-

ﬁcial intelligence to enhance architecture education to

develop digital literacy and holistic competency. Jour-

nal of Artiﬁcial Intelligence in Architecture.

Koh, J. Y., Fried, D., and Salakhutdinov, R. (2023). Gener-

ating images with multimodal language models. 37th

Conference on Neural Information Processing Sys-

tems (NeurIPS 2023).

Li, C., Zhang, T., Du, X., Zhang, Y., and Xie, H. (2024a).

Generative ai models for different steps in architec-

tural design: A literature review.

Li, P., Li, B., and Li, Z. (2024b). Sketch-to-architecture:

Generative ai-aided architectural design. ArXiv,

abs/2403.20186.

Ploennigs, J. and Berger, M. (2023). Ai art in architecture.

AI in Civil Engineering, 2(1).

Roshdy, A., Sharaf, N., Saad, M., and Abdennadher, S.

(2018). Generic data visualization platform. In 2018

22nd International Conference Information Visualisa-

tion (IV), pages 56–57. IEEE.

Sharaf, N., Abdennadher, S., and Fr

uhwirth, T. (2014). Vi-

sualization of constraint handling rules. arXiv preprint

arXiv:1405.3793.

Wan, H., Zhang, J., Chen, Y., Xu, W., and Feng, F.

(2024). Generative ai application for building indus-

try. This work is supported under Contract No. DE-

AC05—76RL01830 with the U.S. Department of En-

ergy.

GRAPP 2025 - 20th International Conference on Computer Graphics Theory and Applications

420