Building the Smart City of Tomorrow: A Bibliometric Analysis of Artificial Intelligence in Urbanization

Abstract

Urbanization is a global trend that continues to grow, leading to an increasing number of people to live in cities. This rapid expansion creates challenges such as traffic congestion, environmental pollution, and the need to ensure high living standards for all residents. To address these challenges, many cities adopt digital technologies to become smarter, more efficient, and more sustainable. Among these technologies, artificial intelligence (AI) has gained significant attention in recent years due to its transformative potential. In the context of smart cities, AI offers innovative solutions across various domains, including mobility, waste management, and energy optimization. Due to its multidisciplinary nature and rapid advancements, research on AI in smart cities has grown significantly. A comprehensive approach is needed to understand its role in urban transformation and identify key research gaps. This paper aims to synthesize existing knowledge on AI in smart cities, providing valuable insights for both researchers and practitioners. We define the scope of AI-related research by analyzing scientific literature and offer three main contributions. First, we provide a holistic overview of the field by conducting a bibliometric analysis to map the status and structure of existing knowledge. Second, we identify major research themes through co-citation clustering. Third, we outline a future research agenda by analyzing the most recent and influential journal articles. Our findings have both theoretical and practical implications for a wide range of disciplines, including computer science, energy, transportation, and security. Furthermore, our results can facilitate collaboration by identifying leading researchers and institutions, highlight critical research gaps, and foster discussions on the benefits and challenges of AI-driven smart city solutions.

1. Introduction

The global population residing in urban areas has increased markedly in recent years [1]. The United Nations projects that by 2030, the urban population will increase by nearly 600 million, reaching a total of 5.2 billion. As of mid-2023, approximately 4.6 billion of the world’s more than 8 billion people resided in urban areas, accounting for 57% of the global population. By 2030, this percentage is expected to rise to 60% [2]. However, urbanization presents several challenges for cities, including traffic congestion, waste management, and increased emissions of carbon dioxide and other greenhouse gases, all of which can negatively impact the quality of life for urban residents [3]. Additionally, cities must address the challenges posed by climate change, as they account for 75% of global natural resource consumption and contribute 60–80% of greenhouse gas emissions [4]. As a result, effective urban growth management is essential to ensuring that citizens have access to a healthy environment, key infrastructure, and essential social services, such as education, healthcare, and housing [5]. To tackle these urban challenges, cities are increasingly adopting innovations and implementing emerging technologies, giving rise to the concept of the “smart city”. The term “smart city” refers to technology-driven innovations within the urban context, including advancements in urban planning and development [6]. Key characteristics of a smart city include a culture of innovation, as well as sustainable, safe, and advanced transportation systems [7].

One group of technologies increasingly explored in smart cities and urban applications is artificial intelligence (AI). AI serves as an umbrella term for various algorithms designed to perform tasks that typically require human intelligence, such as natural language comprehension, pattern recognition, decision-making, and learning from experience [8,9,10]. AI has the potential to enable smart cities to function more autonomously and efficiently, managing urban services and governance in the future [11]. It serves as the foundation for numerous smart city applications, including smart transportation and mobility [12], energy consumption monitoring and optimization [13], and fire hazard risk warnings [14]. Effectively utilizing AI-based techniques can significantly enhance the efficiency and scalability of smart cities and their applications [15]. As a result, AI’s application in smart cities has garnered growing interest from the scientific community, industry, and urban policymakers worldwide. However, the rapid expansion of research on AI in smart cities has led to a lack of comprehensive overviews defining the scope and boundaries of this field, making it challenging for both practitioners and researchers to navigate. Moreover, while AI has its roots in computer science and mathematics, it is inherently interdisciplinary, incorporating contributions from various fields such as economics, neuroscience, and psychology [16,17,18,19]. This multidisciplinary nature further adds to the complexity of AI research in the context of smart cities.

The rapid advancement of AI technologies in smart cities has given rise to a highly dynamic and multidisciplinary research field. AI applications span multiple domains, including education [20], energy [15], health [21], security [22], and transportation [11,23,24]. This convergence of disciplines adds significant complexity to the field. Given the accelerating contributions from diverse academic and practical disciplines, adopting a holistic perspective is crucial. A broad, integrative approach can help clarify how AI is shaping smart city infrastructures, provide insights into the current state of knowledge, and highlight critical research gaps that warrant further exploration. This paper is driven by the need to navigate and synthesize the rapidly evolving body of knowledge on AI in smart cities, aiming to offer researchers and practitioners a comprehensive understanding of the field. In pursuit of this goal, we address the following three research questions.

RQ1: What is the current status and structure of knowledge in the field of AI in smart cities, and what are the key research components and contributing disciplines?

RQ2: What are the major research themes explored in the scientific literature on AI in smart cities?

RQ3: What potential research avenues and themes should be explored in future studies on AI in smart cities?

To address our research questions, we employ a bibliometric analysis. Bibliometric studies quantitatively analyze academic literature, focusing on the measurement and evaluation of scholarly publications. This methodology applies statistical techniques to assess various aspects of research output, including publication trends, citation patterns, authorship dynamics, and the intellectual structure of specific fields [25]. By leveraging a quantitative approach, bibliometric studies provide objective insights and fresh perspectives on a research domain [26,27]. Given our research questions, bibliometric analysis is particularly well-suited for this study. RQ1, which explores the status and structure of AI research in smart cities, benefits from bibliometric techniques such as citation and performance analysis, which reveal key research components, influential works, and contributing disciplines. RQ2, which seeks to identify major research themes, is effectively addressed through keyword-co-occurrence analysis, allowing for the detection of dominant and emerging topics in the field. By offering a large-scale, data-driven, and replicable approach, bibliometric analysis provides an objective and comprehensive understanding of the evolving knowledge landscape in AI and smart cities, making it a robust methodology for addressing these research questions. Additionally, bibliometric studies help uncover the intellectual structure and current state of a field, identifying key topics and future research directions [26,28,29,30]. Similarly, Paul and Criado [31] argue that literature reviews serve as a foundation for future research by synthesizing existing knowledge and highlighting research gaps. In addition to the bibliometric study, we also conducted a literature analysis of the most important research papers from the years 2020 to 2024 to address the third research question.

The structure of this article is as follows: The next section provides essential background on both smart cities and artificial intelligence to establish a foundation for the reader. In the third section, we outline our research methodology, detailing the data collection and analysis process. The fourth section presents the results of our analysis, followed by a proposed future research agenda in the fifth section. Finally, the last section includes a discussion of our findings and concluding remarks.

2. Smart Cities and Artificial Intelligence

To develop a comprehensive understanding of the research at hand, the following sections provide foundational insights into two key areas. The first subsection explores the concepts of smart cities and urbanization, examining how technological advancements are shaping modern urban landscapes. The second subsection delves into artificial intelligence, discussing its principles and its critical role in driving innovation within smart city initiatives.

2.1. Smart Cities and Urbanization

Urbanization is a complex process marked by the shift from agrarian societies to urban-centric living, fundamentally transforming social, economic, and environmental landscapes. Throughout history, cities have served as hubs of attraction, particularly during and after the Industrial Revolution, which significantly accelerated rural-to-urban migration [32]. Thus, urbanization is not a recent phenomenon but has been ongoing for centuries. A key characteristic of urbanization is the spatial expansion of cities, which reflects the growth of urban areas into surrounding rural lands. This expansion leads to shifts in land use patterns, as demonstrated by studies analyzing urban space expansion and its impact on geographical space, highlighting the need for sustainable land-use strategies [33,34]. Among human activities contributing to habitat loss, urban development is one of the primary drivers of local extinction, often causing the disappearance of native species. The rapid expansion of cities and infrastructure not only fragments ecosystems but also disrupts biodiversity, making urbanization one of the most destructive forms of environmental change (see, e.g., [35,36,37]). However, habitat destruction is just one of many challenges associated with expanding cities. One major issue linked to urbanization is air and environmental pollution, largely driven by industrial activity and vehicle emissions. Additionally, the increasing demand for housing and infrastructure results in the loss of green spaces, which play a crucial role in maintaining air quality [38,39]. The effects of rapid urbanization extend to public health, as exposure to pollutants and urban stressors has been linked to rising cases of respiratory illnesses and mental health disorders [40,41,42]. Furthermore, urbanization increases the risk of infectious disease transmission [43], underscoring the need for healthcare to be a central priority in urban planning to mitigate the future impact of such diseases [44].

Technological advancements play a crucial role in driving urbanization. The emergence of smart technologies in cities has given rise to “smart urbanism”, a concept focused on improving urban living conditions through the application of smart technologies [45]. Additionally, the widespread adoption of the internet and digital tools has significantly reshaped urban environments, accelerating city growth and transforming traditional economic and social dynamics [32]. As a result of these developments, the term “smart city” and its related synonyms have gained prominence in both scientific discourse and practical applications. Over the past few decades, the concept of a smart city has been defined in various ways. The term “virtual city” was first introduced by Graham and Aurigi in 1997 [46], highlighting the growing importance of the internet for urban areas. Shortly after, the term “smart city” emerged as a new approach to urban planning [47]. From the late 1990s and early 2000s, the first digital cities began to take shape in Europe and Asia, further sparking significant interest from both the scientific community and industry [48]. Despite its long history, the definition of a smart city has become increasingly ambiguous, as the term is now used in various, sometimes inconsistent, contexts [49]. A comprehensive overview of different definitions of “smart city” is provided by Herath Herath and Mittal [50] and presented in

Table 1. Overview of different definitions of smart cities (based on [50]).

Year	Reference	Definition
2010	Harrison et al. [51]	“[A smart city is] connecting the physical infrastructure, the IT infrastructure, the social infrastructure, and the business infrastructure to leverage the collective intelligence of the city”.
2011	Caragliu et al. [52]	“[A city is smart] when investments in human and social capital and traditional (transport) and modern (ICT) communication infrastructure fuel sustainable economic growth and a high quality of life, with a wise management of natural resources, through participatory governance”.
2014	Galán-García et al. [53] based on Caragliu et al. [52]	“Smart city is a very broad concept which includes not only physical infrastructure but also human and social factors”.
2017	Sterbenz [54]	“[…] cities […] in which ICT (information and communication technology) is deeply embedded all aspects of city operation and service delivery”.
2017	Mustafa and Kar [55]	“The concept of “Smart City” has emerged from “Intelligent Cities”. The basic idea of smart cities is to use the existing resources in a “Smarter” way”.
2019	Cropf and Benton [56]	“[…] cities that significantly incorporate new technologies into their governance”.
2020	Treiblmaier et al. [57]	“A smart city is a geographical area with a high population density that uses information and communication technologies (ICT) to connect and monitor critical infrastructural components and services with the goal of improving the efficiency and the environmental, economic and social sustainability of its operations as well as the quality of life for its citizens”.

below, along with other definitions identified in this research.

Most definitions of smart cities share the common view that the use of digital and information technologies is a key factor in determining a city’s “smartness”. However, a smart city is not solely about employing digital technologies. These technologies should not be used merely for their own sake, but rather with a clear, planned objective. For example, some authors emphasize that the technologies implemented should support the operation and service delivery of a city [54]. Additionally, human and social factors [53] must be considered, alongside the sustainable use of natural resources [52]. Given the comprehensive and holistic nature of this approach, we adopt the definition proposed by Treiblmaier et al. [57]. According to their definition, a smart city is conceptualized as “a geographical area with a high population density that uses information and communication technologies (ICT) to connect and monitor critical infrastructural components and services with the goal of improving the efficiency and the environmental, economic and social sustainability of its operations as well as the quality of life for its citizens” [57] (p. 854).

While ICT serves as a cornerstone for smart cities, the data generated within these urban environments are another critical element. Numerous studies emphasize that data can unlock a wide range of potential benefits for cities [58]. For instance, utilizing open government data [59], analyzing social media [60], and embracing open innovation and crowdsourcing [61] can significantly improve urban life. Moreover, data serves as the backbone for other transformative technologies in smart cities, such as big data analytics [62] and AI [63].

2.2. Artificial Intelligence

The origins of AI can be traced back to 1943, when McCulloch and Pitts proposed the first concept of an artificial neuron [64,65]. Thirteen years later, the term “artificial intelligence” was coined at the Dartmouth Conference [64]. Thus, AI is one of the newest fields explored in science and engineering [66]. Today, it is a complex and dynamic area, encompassing numerous research topics and practical applications for businesses [67]. It is essential to recognize that AI is a multidisciplinary field, involving research in neuroscience, psychology, computer science, and mathematics [18]. In recent years, AI has experienced significant growth and garnered substantial interest from both society and industry, largely driven by advances in computing power and the increasing availability of data for training AI systems [68]. Today, AI refers to a broad array of technologies and approaches applied to a wide range of tasks [16]. Among these, machine learning (ML) is one of the most popular AI methods. ML improves its performance through experience, solving problems using historical data or past examples [69]. ML techniques can be broadly categorized into two main types: supervised and unsupervised learning [70]. While other types of learning exist, such as semi-supervised learning and online learning [71], supervised and unsupervised learning remain the most widely used and preferred [72].

The distinction between supervised and unsupervised learning lies in the presence of labels in the training data [72]. In supervised learning, the system receives labeled examples and input data for training [73]. Broadly speaking, ML involves computational methods that improve performance or make more accurate predictions based on experience [71]. However, the effectiveness of ML and AI algorithms depends heavily on the quality and quantity of their training data [74,75], making these factors crucial [67]. Reinforcement learning is another paradigm that has gained attention in recent AI research. Unlike supervised and unsupervised learning, reinforcement learning involves an AI system, or agent, that learns a behavior to solve a problem through trial-and-error interactions with its environment [76]. In this approach, AI systems effectively become their own teachers, learning without human-provided data, guidance, or knowledge [77].

In recent years, AI has become a transformative force in both business and society. AI now powers many aspects of modern life, such as identifying objects in images, a feature increasingly common in consumer products like smartphones and cameras [78]. For companies, AI technologies have a broad range of potential use cases and application areas. For instance, AI technologies enable the creation of innovative business models [79], while also allowing for the automation of routine tasks, helping companies focus on more strategic initiatives. For example, AI-powered automation in customer service, such as chatbots, enables businesses to provide 24/7 support, lowering operational costs while boosting customer satisfaction [80,81]. Additionally, AI’s ability to analyze large datasets can enhance decision-making, optimize operations, and develop strategic and competitive advantages [82,83]. Finally, forecasting is one of the use-cases AI and ML techniques are most frequently used for within companies [16]. Examples include the application to forecast profit and loss variables, such as revenues [84], as well as cash flow variables, like customer payment dates [85].

AI methods can be used for a wide range of applications beyond the scope of businesses. In medicine, for example, AI is employed in areas like precision and personalized medicine [86,87] and in cancer detection [88]. Within smart cities, AI serves as the foundation for a variety of applications aimed at improving urban life. One central use case is enhancing urban livability and controlling air pollution. By integrating AI with environmental monitoring systems and IoT sensors, cities can leverage real-time data analytics to assess air quality and identify pollution sources more effectively. This combination of IoT and AI enables data-driven solutions for pollution control, significantly improving urban sustainability and livability [89]. For example, AI algorithms can forecast air quality indices by analyzing meteorological data and pollution sources, allowing for timely interventions to reduce air pollution [90,91,92]. Additionally, AI-driven models can support urban planning by simulating how different urban configurations impact air quality, promoting sustainable development [93]. AI technologies also form the foundation for AI-based monitoring systems in smart buildings (AIMS-SB), which help manage energy consumption and optimize energy production and recycling [94]. Thus, integrating AI into smart city initiatives not only improves urban livability by fostering healthier environments but also provides innovative solutions for air pollution control. However, it is important to acknowledge that while AI applications in smart cities offer advantages like automation and increased efficiency, they also bring regulatory concerns, including the potential for discrimination in service delivery, as well as issues related to privacy, legal implications, and ethical challenges [50,95].

3. Research Method

In this section, we describe the bibliometric approach used in our study, which consists of two main steps: data collection and data analysis. The first subsection outlines the data collection process, explaining how we obtained the bibliometric data from Scopus. Following data collection, the second step involves data analysis, which is described in the second subsection, including the methods and tools utilized.

3.1. Data Collection

The first step of our bibliometric study was the collection of bibliometric data. There are several databases available, each with distinct features and functionalities [96]. Among the most widely used are Scopus and Web of Science [97,98]. Although different databases do exist, most of the bibliometric studies that are tool-supported rely only on one database [99], since most of the bibliometric software is not able to integrate data from different sources [100,101]. Hence, we followed the recommendation of Donthu et al. [25] and opted to gather our bibliometric data from a single database.

Among the available databases, we chose Scopus for our data collection. Scopus is a database owned by Elsevier that allows retrieval of scholarly articles via a text-based search query [100,102]. There are several reasons for this choice. First, Scopus is a widely recognized and extensively used database, having been utilized in numerous recent bibliometric studies (e.g., [88,97,103,104,105]). Compared to Web of Science (WoS), Scopus covers a broader range of journals and publications published after 1990 [106,107,108], making it particularly suitable for timely topics like AI within smart cities. Additionally, Scopus’ reliable management of cited works and curated indexing make it a more dependable resource than Google Scholar [105].

In developing our search string, we took care to include all key and common synonyms to ensure comprehensive coverage of the relevant literature. Our search string consists of two parts: the technology (AI) and the domain (smart city). For the AI part, we consulted recent bibliometric studies [88,109,110,111] to identify important synonyms, including machine learning, deep learning, and evolutionary computation. Similarly, to identify synonyms for smart city, we reviewed other recent bibliometric studies and literature reviews in this area [107,112,113,114,115]. Based on this, we included additional terms such as “intelligent city”, “digital city”, and “smart sustainable city”. This resulted in the following search string, which was used for literature collection in both titles and abstracts:

((“artificial intelligence” OR “machine intelligence” OR “autonomous robot*” OR “autonomous agent*” OR “intelligent robot*” OR “intelligent agent*” OR “artificial neural network*” OR “Machine learn*” OR “Deep learn*” OR “thinking computer system” OR “fuzzy expert system*” OR “evolutionary computation” OR “hybrid intelligent system*”) AND

(“smart city” OR “smart cities” OR “intelligent city” OR “intelligent cities” OR “digital city” OR “digital cities” OR “smart sustainable city” OR “smart sustainable cities”))

The search was conducted on 19 September 2024, resulting in an initial set of 6154 identified documents. To refine our sample, we performed several exclusion steps. First, we limited our search to journal articles, conference papers, and reviews, excluding other document types such as book chapters, editorials, and short surveys. This step led to the removal of 1263 articles, leaving 4891 documents. Subsequently, we further restricted the selection to include only final-published articles in English. After excluding non-English and non-final articles, as well as articles with no author information, 4724 documents remained, which were then imported as a CSV file. To ensure data quality, we conducted a manual data-cleaning process. We checked for and removed duplicate and erroneous entries by hand. However, we chose not to filter papers based on perceived relevance, as “smart city” is a broad and interdisciplinary topic. Applying subjective filtering criteria could introduce bias and limit the comprehensiveness of our analysis. After completing these steps, the final sample consisted of 4719 documents.

3.2. Data Analysis

There is a wide array of software and tools available to support the implementation of bibliometric studies, each with its own features, strengths, and limitations (for a comprehensive overview, see [96]). For this study, we used a combination of two tools: Biblioshiny/Bibliometrix and VOSviewer. First, Bibliometrix is an open-source R package developed by Aria and Cuccurullo [116] that offers a wide range of analytical tools for bibliometric data [96]. We complemented Bibliometrix with Biblioshiny, which provides a user-friendly interface for easier visualization of bibliometric data [96]. In addition, we supplemented Biblioshiny/Bibliometrix with VOSviewer, a tool developed by the Centre for Science and Technology Studies at Leiden University in the Netherlands, specifically designed for the visualization of bibliometric data [96,117,118]. In our research, we used VOSviewer to conduct bibliographic coupling analysis and to construct a keyword co-occurrence network.

4. Findings

The following section presents the findings of our bibliometric study. In the first subsection, we provide an overview of the research field, including general metrics, annual publication numbers, and the distribution of articles across research disciplines. The second subsection offers a performance analysis of affiliations, funding institutions, and publication outlets. The third subsection focuses on countries and their collaborative efforts. Finally, the last subsection presents a thematic analysis, featuring a keyword co-occurrence network and the most frequently occurring keywords.

4.1. Overview and General Metrics

We begin by presenting general information about our dataset (see

Table 2. Overview of general metrics of the data sample.

Metric	Value
Main information
Timespan of the sample	2006–2024
Sources and outlets	1875
Documents	4719
Average citations	17.26
Annual growth rate	44.27 percent
Average age of a document	3.47 years
Number of author’s keywords	9752
Document Types
Journal article	2310
Conference paper	2182
Review	227
Authors
Number of different authors	13,081
Documents per author	0.36
Authors of single-authored documents	308
Authors of multi-authored documents	12,773
Single-authored documents	328
Multi-authored documents	4391
Collaboration index	2.91

and Figure 1). The 4719 documents identified span from 2006 to 2024, covering 1875 different sources and outlets. The documents have an average of 17.24 citations, with an annual growth rate of 44.28%. On average, the documents are 3.47 years old, and there are 9752 unique author keywords in the dataset. The majority of the documents are journal articles (2310), followed by conference papers (2182), and reviews (227). A total of 13,081 different authors contributed to the dataset, which equates to an average of 0.36 documents per author. Among these, 308 authors published single-authored documents, while 12,773 authors contributed to multi-authored works. The dataset includes 328 single-authored and 4391 multi-authored documents, resulting in a collaboration index of 2.91 (calculated as 12,773 divided by 4391).

Figure 2 displays the annual number of publications of the articles within our sample. The data show a significant increase in the number of publications, particularly from 2017 onwards, suggesting a growing interest in this research area. The first article in our sample is a conference paper that was published in 2006 and is entitled “The architecture of intelligent cities: Integrating human, collective and artificial intelligence to enhance knowledge and innovation” [119]. From 2006 to 2016, the number of publications remained relatively low, ranging from 0 to 40 articles per year. The field started to gain noticeable attention in 2017, with 77 publications, followed by a sharp rise in 2018 (215 publications) and 2019 (357 publications). This upward trend continued, with 553 publications in 2020 and 705 in 2021. The most significant increase occurred in 2018 (215 publications), with almost three times as many publications as in the previous year (77 publications). In 2023, the number of articles peaked at 1105. In 2024, although the year is not yet complete, there have already been 740 articles published by the date of our data collection (19 September 2024). If we assume the same publication rate throughout the rest of the year, we can expect around 290 additional papers to be published by the end of the year. This suggests that the upward trend may either continue or begin to stabilize. The overall growth over the examined period reflects the increasing importance and integration of AI in smart city developments, as well as the corresponding rise in research interest and activity within this domain.

Figure 3 shows the distribution of research on AI in smart cities. The data in Figure 2 was sourced from Scopus, where publications are categorized into disciplines based on the outlet in which they appear. However, certain journals or conferences may be associated with multiple disciplines. As we can see, the research within this area spans across many different disciplines. The largest portion of the research is contributed by the field of Computer Science, which represents 33% of the total. This is followed by Engineering, which accounts for 22%, reflecting the crucial role of technological innovation and computational systems in the development and implementation of AI solutions in urban environments. Mathematics contributes 8% of the research, emphasizing the importance of algorithmic development and data analysis in AI applications (see, e.g., [120,121,122]). Social Sciences, with a 6% share, highlights the growing interest in understanding the societal implications of AI integration in cities, including ethical considerations and public (see, e.g., [123,124]). The fields of Decision Sciences and Business and Management each account for 5% of the total, demonstrating the significance of strategic decision-making processes and the business models that support AI-driven urban [125,126]. The same percentage (5%) is attributed to Energy, reflecting research efforts aimed at optimizing resource management and sustainable energy systems through AI. Smaller but notable contributions come from Materials Science (2%), Environmental Sciences (5%), and Medicine (2%), which showcases the interdisciplinary nature of AI in addressing specific challenges within smart cities, such as environmental monitoring, public health, and infrastructure resilience. Additionally, the category labeled “Others” accounts for 7%, representing a variety of other disciplines involved in the research on AI in urban settings. In summary, the figure highlights the multidisciplinary nature of AI research in smart cities, with a predominant focus on technological and computational domains, while also recognizing the growing relevance of social and environmental aspects. This diverse engagement suggests that a holistic approach is being adopted to address the complex challenges associated with smart city development.

4.2. Performance Analysis

In this section, we will provide some information about the most productive affiliations, the most relevant funding sources, as well as the most important outlets. First,

Table 3. The 15 most productive affiliations contributing to research on AI in smart cities.

Rank	Affiliation	Country	Publications
01	King Saud University	Saudi Arabia	64
02	Prince Sattam Bin Abdulaziz University	Saudi Arabia	56
03	King Abdulaziz University	Saudi Arabia	53
04	Princess Nourah Bint Abdulrahman University	Saudi Arabia	51
05	Vellore Institute of Technology	India	41
06	King Khalid University	Saudi Arabia	39
07	Chinese Academy of Sciences	China	34
08	Wuhan University	China	34
09	Qatar University	Qatar	34
10	SRM Institute of Science and Technology	India	33
11	University of Petroleum and Energy Studies	India	32
12	Saveetha Institute of Medical and Technical Sciences	India	31
13	Sejong University	South Korea	30
14	K L Deemed to be University	India	30
15	The Hong Kong Polytechnic University	Hong Kong	29

highlights the most productive affiliations contributing to research on AI in smart cities. Among the top institutions, Saudi Arabian universities dominate, with King Saud University leading the list with 64 publications, followed by Prince Sattam Bin Abdulaziz University and King Abdulaziz University, with 56 and 53 publications, respectively. Princess Nourah Bint Abdulrahman University and King Khalid University also make notable contributions, reflecting Saudi Arabia’s substantial investment in smart city and AI research. India is well represented by Vellore Institute of Technology (41 publications), SRM Institute of Science and Technology (33 publications), and other institutions ranking highly. Chinese institutions, such as the Chinese Academy of Sciences and Wuhan University, also play a significant role, each contributing 34 publications. Qatar University, Sejong University (South Korea), and Hong Kong Polytechnic University complete the list, indicating that Asia and the Middle East are key regions driving research in this domain forward. The strong presence of Saudi Arabian universities highlights the country’s commitment to fostering innovation in AI and smart cities, while the participation of institutions from India, China, and other parts of Asia underscores the global interest in this field. These affiliations contribute significantly to the knowledge base, offering diverse perspectives on smart city technologies and their applications.

Table 4. The 15 most relevant funding sponsors of research on AI in smart cities.

Rank	Affiliation	Country/Region	Funded Articles
01	National Natural Science Foundation of China	China	313
02	European Commission	European Union	103
03	National Science Foundation	United States	88
04	National Key Research and Development Program of China	China	84
05	King Saud University	Saudi Arabia	83
06	Ministry of Science and Technology of the People’s Republic of China	China	82
07	National Research Foundation of Korea	South Korea	71
08	Horizon 2020 Framework Programme	European Union	69
09	European Regional Development Fund	European Union	66
10	Ministry of Science, ICT and Future Planning	South Korea	51
11	Fundamental Research Funds for the Central Universities	China	46
12	Fundação para a Ciência e a Tecnologia	Portugal	42
13	Natural Sciences and Engineering Research Council of Canada	Canada	36
14	Institute for Information and Communications Technology Promotion	South Korea	34
15	Ministry of Education of the People’s Republic of China	China	31

provides insights into the most influential funding sponsors contributing to research in AI and smart cities. Topping the list is the National Natural Science Foundation of China, with 313 publications. Other Chinese institutions, such as the National Key Research and Development Program of China (84 publications) and the Ministry of Science and Technology of China (82 publications), further signify China’s dominant role in funding research efforts in this domain. Together with the presence of several Chinese universities and programs in the previously discussed affiliation rankings, as well as the country’s dominant role (see below), this reflects the country’s strategic prioritization of AI and smart city development. Ranking second is the European Commission, with 103 publications. Other programs like the Horizon 2020 Framework Programme (69 publications) and the European Regional Development Fund (66 publications) show the European Union’s commitment to advancing AI research and demonstrate the EU’s emphasis on collaborative, cross-border research efforts. The National Science Foundation (88 publications) from the United States also appears as a significant contributor. Notably, King Saud University is the only individual university among the top 15 funders, with 83 funded publications, highlighting Saudi Arabia’s active investment in this field. South Korea, through its National Research Foundation (71 publications) and the Ministry of Science, ICT, and Future Planning (51 publications), also demonstrates a strong commitment to research. Other significant contributors include Fundação para a Ciência e a Tecnologia from Portugal and the Natural Sciences and Engineering Research Council of Canada.

Table 5. The 15 most important outlets on AI in smart cities.

Rank	Outlet	Articles
01	IEEE Access	152
02	Lecture Notes In Networks And Systems	101
03	Sensors	96
04	Sustainability (MDPI)	77
05	Lecture Notes In Computer Science	75
06	IEEE Internet Of Things Journal	72
07	ACM International Conference Proceeding Series	71
08	Applied Sciences (MDPI)	59
09	Sustainable Cities And Society	52
10	Communications In Computer And Information Science	51
11	Smart Cities	49
12	Electronics (MDPI)	49
13	Lecture Notes In Electrical Engineering	45
14	Sensors (MDPI)	36
15	Procedia Computer Science	36

highlights the top 15 journals and conference proceedings that have published the most research on AI in smart cities. IEEE Access leads the list with 152 articles, followed by Lecture Notes in Networks and Systems with 101 articles, indicating the importance of research on network technologies that form the backbone of smart city infrastructure. Similarly, Sensors (96 articles) and Sustainability (MDPI) (77 articles) emphasize the growing intersection of AI, sensor technology, and sustainable urban development. Other prominent outlets include Lecture Notes in Computer Science (75 articles) and the IEEE Internet of Things Journal (72 articles), both of which focus on computing and IoT technologies that are essential for the intelligent functioning of smart cities. Notably, Sustainable Cities and Society (52 articles) reflects the importance of sustainability in smart city research, while Smart Cities (49 articles) focuses specifically on this emerging area of urban development. The wide range of outlets, from Electronics (MDPI) to Applied Sciences (MDPI), underlines that research in AI and smart cities spans multiple disciplines, including electronics, computer science, urban studies, and sustainability. This interdisciplinary approach seems to be key to solving the complex challenges that smart cities face today.

4.3. Geographic Distribution of Research Contributions

In the following, we will take a closer look at the contributing countries within our sample, as well as the collaborative structure of the countries. We start by giving an overview of the 15 highest-contributing countries in terms of total article count in

Table 6. Overview of the 15 most productive countries in terms of total article count.

Rank	Country	Articles	First Publication	Citations	Citations per Document	Annual Growth Rate
01	India	1055	2015	13,938	13.21	80.56%
02	China	942	2007	19,057	20.23	34.64%
03	United States	525	2012	16,946	32.28	41.05%
04	Saudi Arabia	370	2017	8216	22.21	56.18%
05	United Kingdom	252	2012	8760	34.76	34.16%
06	South Korea	242	2011	7537	31.14	28.10%
07	Australia	195	2015	10,534	54.02	28.67%
08	Italy	192	2013	3073	16.01	33.99%
09	Canada	184	2017	4422	24.03	47.24%
10	Pakistan	172	2016	4684	27.23	49.53%
11	Spain	146	2011	2679	18.35	21.81%
12	UAE	143	2018	3317	23.20	50.24%
13	Malaysia	134	2017	2886	21.54	48.60%
14	Egypt	122	2017	2470	20.25	48.60%
15	Germany	104	2012	2403	23.11	20.09%

. India ranks first with 1055 articles since its first publication in 2015. Despite being a relatively recent entrant, India’s contribution has grown significantly, achieving an impressive annual growth rate of 80.56%, which is the highest among the top 15 countries. India’s research has accumulated 13,938 citations, with an average of 13.21 citations per document. China is the second-most productive country, with 942 publications starting from 2007. Chinese publications have garnered the highest number of citations (19,057), with an average of 20.23 citations per document. Despite a lower growth rate compared to India (34.64%), China’s earlier start in the field suggests its sustained influence. The United States ranks third, with 525 articles published since 2012. Despite producing fewer publications than India and China, U.S. research has one of the highest average citations per document at 32.28, totaling 16,946 citations. The annual growth rate of U.S. research stands at 41.05%.

Figure 4 illustrates the global distribution of country-specific research output related to AI in smart cities. Countries shaded in blue represent those actively contributing to this field, with darker shades indicating higher levels of publication activity. Along with Table 4 above, nations such as the United States, China, India, and several European countries show a strong presence in this domain, reflecting their substantial investments in AI research and smart city initiatives. In contrast, many regions in Africa, parts of Central America, and some areas in Asia are depicted in gray, indicating lower levels of research output or no articles in our sample. This disparity highlights the uneven global distribution of AI research, likely due to variations in economic development, research infrastructure, and governmental support for innovation in urban technology. The concentration of research activity in developed countries underscores the importance of addressing the gap between emerging and established economies. Collaborative efforts and international partnerships could help promote AI-driven smart city solutions more equitably, fostering global advancements in urban sustainability, resource management, and quality of life improvements.

Finally, in Figure 5, we show a country collaboration network. A country collaboration network is a graphical representation of the collaborative relationships between countries in a given field of research. In the context of scientific studies, such a network typically shows how often researchers from different countries work together on publications. The nodes in the network represent countries, and the edges or links between them indicate collaboration, usually based on co-authorship in scientific articles. Countries that have high centrality tend to collaborate with many other countries and may act as “hubs” connecting different parts of the network. Prominent nations such as the United States, China, India, and the United Kingdom are among the most central and connected nodes, suggesting that they are the most actively involved in international collaborations on AI and smart cities. These countries, due to their significant research output and extensive global partnerships, serve as key hubs in the network. The United States and China, positioned as central players, maintain extensive ties with countries around the world, showcasing their leading role in shaping the direction of AI and smart city research on a global scale. Additionally, regional leaders such as Saudi Arabia and India exhibit strong collaborative ties within their respective clusters, demonstrating their growing influence in this area of research, particularly in the Middle East and South Asia.

4.4. Thematic Analysis

In this section, we will dive into the thematic areas and topics that are addressed in research on AI in smart cities. We start by showing the most frequently occurring keywords within our sample that were derived from Scopus (

Table 7. Overview of the 30 most frequently occurring keywords (based on Scopus).

Rank	Keyword	Amount
01	Internet of Things/IoT	1437/323
02	Learning Systems	900
03	Learning Algorithms	479
04	Forecasting	425
05	Network Security	368
06	Big Data	345
07	Intelligent Systems	305
08	Decision Making	280
09	Automation	259
10	Convolutional Neural Networks/Convolutional Neural Network	254/218
11	Information Management	215
12	Traffic Congestion	214
13	Energy Utilization	212
14	Edge Computing	206
15	Sustainable Development	198
16	Long Short-term Memory	194
17	Energy Efficiency	190
18	Intelligent Buildings	185
19	Object Detection	180
20	Convolution	177
21	Digital Storage	175
22	Decision Trees	175
23	Blockchain	174
24	5G Mobile Communication Systems	174
25	Intrusion Detection	171
26	Security	167
27	Intelligent Transportation Systems	164
28	Data Handling	164
29	Security Systems	161
30	Data Mining	159

). Furthermore, the Figure 6, Figure 7 and Figure 8 display word clouds of the most frequently occurring author keywords in our sample for different time spans. For the creation of Table 7 and Figure 6, Figure 7 and Figure 8, we excluded keywords that were included in our search query, such as “smart city” or “artificial intelligence”. As can be seen from the table and all word clouds, the “Internet of Things” (IoT) is one of the most frequently addressed topics in AI and smart city research. This can be explained by the pivotal role that IoT plays in the development and functionality of smart cities by enhancing urban management and improving the quality of life for residents. IoT facilitates the interconnection of various devices and systems, enabling real-time data collection and analysis across multiple urban sectors, including transportation, energy, and waste management (see, e.g., [127,128,129]). Not surprisingly, IoT is also one of the most dominant terms in all three word clouds. Energy-related topics are also significant within the analysis, with “Energy Utilization” (212 mentions) and “Energy Efficiency” (190 mentions) reflecting the growing emphasis on optimizing energy consumption in smart cities. Achieving sustainable development (198 mentions) is a critical goal for urban planners, and smart city technologies play a vital role in reducing energy waste and promoting green infrastructure.

The following figures present word clouds with the most frequently occurring keywords during different periods. The first word cloud covers the years 2006 to 2019 and represents the foundational phase, during which early research explored the initial integration of AI into smart cities (SCs). The second phase (2020–2022) reflects a period of rapid development, in which advancements in AI technologies and their applications in SCs gained significant momentum. Finally, the last word cloud highlights the most recent trends (2023–2024), offering a glimpse into emerging directions and cutting-edge innovations shaping the future of the field.

First, Figure 6 shows the most prominent and frequently occurring author keywords for the earliest period (2006–2019). The prominence of “Big Data” reflects the early emphasis on managing the vast quantities of data generated by smart city systems. This involves applications in different areas as well as the consideration of challenges such as security issues and ethical considerations (see, e.g., [130,131]). Keywords such as “Classification”, “Security”, and “Object Detection” point to the growing use of AI and machine learning techniques to enhance the efficiency and safety of urban systems. Security has become a major focus due to rising concerns over cybersecurity threats, data privacy, and the need for AI-driven surveillance and risk detection in urban spaces. Object Detection is gaining prominence as smart cities increasingly deploy AI-powered computer vision for tasks like automated traffic control, pedestrian monitoring, and infrastructure maintenance. The growing complexity of urban systems and the push for automation, efficiency, and safety drive the relevance of these AI applications.

In the later period (2020–2022, see Figure 7), there are some notable shifts. While “Big Data”, “IoT”, and “Cloud Computing” remain highly relevant, new terms like “Federated Learning”, “Sustainability”, and “Anomaly Detection” have gained prominence, reflecting a maturing field. “Federated Learning” represents a shift toward privacy-preserving, decentralized AI models, particularly crucial in smart cities where data security and privacy are paramount. The rise of “Sustainability” highlights a growing focus on using AI to address environmental concerns and enhance energy efficiency in urban systems. As urban populations grow, cities face mounting pressure to reduce carbon emissions, improve air quality, and enhance energy efficiency. AI-driven solutions, such as smart grids, predictive analytics for energy consumption, and AI-powered environmental monitoring, help optimize resource management and promote eco-friendly urban planning. Additionally, AI supports climate adaptation strategies by predicting environmental risks, optimizing waste management, and improving transportation efficiency to reduce emissions. This growing emphasis on sustainability reflects the urgent need for smarter, data-driven approaches to create greener, more resilient cities. While “Security” was already important in 2006–2019, it became even more central in 2020–2022, possibly due to the growing complexity of urban infrastructures and the increasing reliance on digital systems vulnerable to cyberattacks.

Finally, Figure 8 depicts a word cloud of the latest years within our sample (2023 and 2024). The word cloud for 2023–2024 reflects a shift towards more secure, privacy-focused, and sustainable AI applications in smart cities, driven by innovations like Federated Learning and robust cybersecurity solutions. This shows that smart city research is now focusing not only on technical advancements but also on ensuring safety, privacy, and sustainability in an increasingly interconnected urban world. Compared to previous periods, “Security” has moved from a secondary to a primary focus in recent years. While it was always present, the 2023–2024 word cloud shows an intensified concern for safeguarding smart city technologies from potential breaches. Furthermore, it is noteworthy that blockchain technology, although present in the earlier periods as well, seems to be of growing interest. Blockchain refers to a group of technologies that can be defined as “[…] digital, decentralized, and distributed ledger[s] in which transactions are logged and added in chronological order to create permanent and tamper-proof records” [132]. Blockchain has several applications in areas related to smart cities [133], including use cases in smart mobility [134,135], tourism [136], and healthcare [137,138]. The convergence of AI and blockchain is increasingly being investigated (see, e.g., [139,140,141]) and involves applications such as increased security for IoT applications [142] and traffic flow prediction in urban computing [143].

4.4.1. Complementing Technologies and Security

After we identified the most frequently occurring keywords and their prevalence across different periods, we now analyze the correlations between these keywords to gain deeper insights. Figure 9 presents a keyword co-occurrence network of the most frequently appearing author keywords (with a minimum occurrence of 20). The size of each circle and the font size of the keywords reflect their frequency: the more frequently a keyword appears, the larger its representation. This principle also applies to the lines connecting the keywords—the more often two keywords are mentioned together in the same publication, the thicker the line connecting them. Additionally, keywords that frequently co-occur are grouped into clusters represented by the same color. In total, five distinct clusters can be observed.

The red cluster is by far the largest and most prominent thematic area revealed by the keyword co-occurrence network. It contains central terms such as “Internet of Things”, “Big Data”, “Data Mining”, and “Wireless Sensor Networks”, reflecting the fundamental technologies that underpin smart cities and their intersection with AI. These terms are closely linked because AI plays a crucial role in enhancing their capabilities, such as improving real-time data processing, automation, and decision-making One notable example is the convergence of Blockchain, IoT, and AI, which has been explored for secure and decentralized smart city applications (see, e.g., [141,144,145]). Additionally, this cluster includes key AI methods such as “Classification”, “Clustering”, as well as their application in areas like “Industry 4.0”. Another significant theme within this cluster is “Security”, which frequently appears alongside related terms. Researchers in this domain investigate how AI can enhance security in smart city environments, particularly in cybersecurity and intrusion detection within IoT-based infrastructures ([146,147,148,149]. The presence of terms such as “Federated Learning” and “Edge Computing” suggests ongoing efforts to develop more secure, decentralized AI-driven systems.

It is important to note that the red cluster is not an isolated system but rather a collection of core concepts that extend into various domains and interact with other clusters. A closer examination would likely reveal distinct subtopics, such as AI-driven network security, anomaly detection in smart grids, and privacy-preserving AI applications. These findings highlight the central role of AI in structuring and advancing smart city technologies while also pointing to emerging interdisciplinary research directions.

4.4.2. Intelligent Transportation and Smart Mobility

The green cluster comprises terms related to traffic, transportation, and urban mobility, reflecting a significant area of AI applications in smart cities. This cluster includes research investigating how AI technologies can optimize mobility systems, enhance efficiency, and improve overall transportation management in urban environments. The applications that are investigated are manifold. For example, ref. [150] analyze how AI can be used for intelligent traffic forecasting, while [151] propose deep reinforcement learning for adaptive traffic light control. Other examples include AI applications for parking management systems [152], traffic prediction [153], vehicle identification and counting [154], and AI-driven traffic flow management [155]. Beyond mobility-related topics, the green cluster also includes terms such as “Prediction”, “Air Quality”, and “Air Pollution”, indicating a strong connection between AI-driven transportation research and environmental monitoring. Studies in this area apply deep learning and federated learning techniques to predict and forecast air quality [156,157], contributing to more effective pollution control and urban planning strategies. Furthermore, AI-driven environmental monitoring systems have been explored to track and assess pollution levels in real time [158], underscoring the growing interdisciplinary link between transportation, sustainability, and AI.

This cluster highlights the crucial role AI plays in shaping the future of intelligent urban mobility. It not only enables more efficient traffic and transportation management but also integrates with broader environmental concerns, emphasizing the need for AI-driven solutions that balance efficiency with sustainability. As research in this domain continues to evolve, emerging trends such as edge computing for real-time traffic analytics and AI-enhanced multimodal transportation systems are likely to further enrich the field.

4.4.3. AI-Based Energy Efficiency

AI plays a crucial role in energy management within smart cities by leveraging advanced technologies to optimize energy consumption, enhance sustainability, and improve urban living standards. The yellow cluster in the keyword co-occurrence network focuses on this domain encompassing key terms such as “Energy Consumption”, “Energy Management”, and “Smart Building”. Research in this area explores how AI can enable more efficient energy management by integrating predictive analytics, machine learning, and automation into urban energy systems. One of AI’s most significant contributions to this field is optimizing the use of renewable energy sources, which helps reduce carbon emissions and improve overall energy efficiency [159,160]. Additionally, AI enhances the operational efficiency of smart grids, which are fundamental to modern urban energy management. Through real-time data processing and machine learning algorithms, AI-driven smart grids can dynamically adjust to fluctuations in energy supply and demand, ensuring a more stable and efficient distribution network [161,162].

Beyond large-scale energy infrastructure, AI technologies also contribute to sustainable energy practices at the building level. Smart buildings equipped with AI-powered systems can predict energy demand, optimize heating and cooling processes, and improve renewable energy integration [163], For instance, ref. [163] demonstrate how AI can support energy efficiency by forecasting renewable energy production and evaluating energy recycling strategies [94]. These advancements highlight the role of AI in enabling data-driven decision-making for sustainable urban development, further reinforcing its significance in the evolution of smart cities.

As research in this area progresses, emerging trends such as AI-driven demand response management, decentralized energy trading via blockchain, and AI-enhanced IoT integration in energy networks are likely to further transform urban energy systems. The yellow cluster thus represents a critical intersection of AI, sustainability, and smart infrastructure, offering valuable insights into the future of intelligent energy management in cities.

4.4.4. Computer Vision and Object Detection

The blue cluster in the keyword co-occurrence network focuses on object detection and computer vision, highlighting their diverse applications within smart cities. AI-powered object detection technologies are increasingly being explored for a range of urban challenges, including video-based vehicle counting to optimize traffic flow [164], fire detection for improved emergency response systems [165], and various smart traffic management solutions [166]. These applications demonstrate the crucial role of AI-driven computer vision in enhancing urban infrastructure and public safety.

One of the most extensively researched topics within this cluster is waste management, a critical issue in the context of rapid urbanization and rising waste generation. AI systems significantly improve waste management in smart cities by automating waste sorting and classification, thereby increasing efficiency and reducing environmental impact. Machine learning and deep learning techniques enable AI models to accurately identify and categorize waste materials, leading to more effective recycling and disposal processes. For instance, ref. [167] that AI systems are able to identify and sort waste with an accuracy ranging from 72.8% to 99.95%. Furthermore, AI applied to waste logistics can lead to a reduction in transportation distances of up to 36.8%, a decrease in costs of up to 13.35%, and time savings of up to 28.22% [167].

The blue cluster underscores the expanding role of AI-powered computer vision in smart city development. From traffic optimization to environmental sustainability, the integration of object detection technologies continues to drive innovation in urban management. Future research in this area is likely to explore advancements in real-time surveillance analytics, AI-enhanced public safety monitoring, and fully autonomous waste sorting systems, further solidifying the importance of AI in building cleaner, more efficient cities.

4.4.5. Governance and Urban Planning

The light blue cluster is the smallest within the keyword co-occurrence network, comprising four key terms: “Sustainability”, “Governance”, “Urban Planning”, and “Remote Sensing”. Despite its size, this cluster is highly interconnected with other thematic areas, particularly the yellow cluster, which focuses on AI-driven energy efficiency (Section 4.4.3), and the blue cluster, which examines computer vision applications, including waste management (Section 4.4.4). These connections highlight the interdisciplinary nature of AI research in smart cities, where sustainability, governance, and urban planning intersect with technological advancements in energy management and environmental monitoring.

At the core of this cluster is the concept of “Sustainability”, which serves as a bridge between various AI applications aimed at optimizing urban systems. AI-driven predictive analytics contribute to sustainability by enhancing energy efficiency, enabling real-time adjustments in energy consumption, and minimizing resource waste [168,169]. Additionally, AI-powered traffic management systems play a crucial role in urban sustainability by reducing congestion and emissions, thereby improving air quality and overall mobility [170]. What sets the light blue cluster apart, however, is its emphasis on the relationship between AI, governance, and urban planning. Researchers have explored how AI can be leveraged to predict urban expansion and evolution, aiding city planners in making data-driven decisions for sustainable urban growth [171]. AI has also been applied in urban design, where it helps optimize spatial planning and infrastructure development by analyzing large-scale urban data or [172]. Beyond planning, an important area of research within this cluster concerns the governance of AI in smart cities. Scholars examine how AI-driven systems should be regulated and managed to ensure ethical, transparent, and responsible deployment [173]. This includes discussions on policy frameworks, AI accountability, and the potential risks and benefits of AI integration in urban governance [174].

Although small, the light blue cluster represents a critical intersection between AI, sustainability, and policy-making, emphasizing the need for governance models that balance innovation with societal and environmental concerns. Future research in this area is likely to delve deeper into AI-assisted policy design, ethical AI frameworks for urban governance, and the role of AI in achieving long-term sustainability goals.

5. Future Research

In this section, we aim to present avenues for research with the goal of helping facilitate future research and providing a framework for upcoming topics for both practitioners and researchers. To derive our future research agenda, we qualitatively analyzed the most cited articles from the last four years (2020–2024). Because the total number of publications is far too high for a qualitative analysis, the publications were ranked by normalized total citations (NTC) using Bibliometrix. NTC is used to compare publications of different ages, subject matters, and document types to identify the most impactful ones. To calculate NTC, the citations of a publication are divided by the expected citation rate for a document published in a given year, of a specific type, and within a certain domain. Therefore, NTC was used to create an unbiased ranking of articles published in recent years. The 40 articles with the highest NTC score (see

Table 8. List of the 40 most relevant articles published between 2020 and 2024 (based on their NTC).

Rank	NTC	Article
01	78.86	Sarker [175]
02	71.36	Bibri et al. [176]
03	35.86	Fuller et al. [177]
04	34.83	Habbal et al. [178]
05	32.56	Gad [179]
06	28.84	Talaat and ZainEldin [165]
07	22.31	Nguyen et al. [180]
08	20.22	Cao et al. [181]
09	19.52	Sarker et al. [182]
10	19.44	Dargan et al. [183]
11	19.02	Loh et al. [184]
12	18.99	Pandya et al. [185]
13	18.48	Allam et al. [186]
14	18.27	Javed et al. [187]
15	16.00	Alahi et al. [188]
16	15.29	Ullah et al. [189]
17	15.23	Malekloo et al. [190]
18	14.54	Shi et al. [191]
19	14.35	Li et al. [192]
20	14.24	Javed et al. [193]
21	14.15	Saleem et al. [194]
22	13.89	Oladimeji et al. [195]
23	13.60	Ghazal et al. [21]
24	13.59	Fadhel et al. [196]
25	13.33	Akhter and Sofi [197]
26	13.33	Ullah et al. [15]
27	13.06	Otoum et al. [198]
28	12.74	Kalapaaking et al. [199]
29	12.74	Hijazi et al. [200]
30	12.74	Kurunathan et al. [201]
31	12.12	Singh et al. [202]
32	11.96	Li et al. [203]
33	11.78	Rani and Sharma [204]
34	11.35	Singh et al. [141]
35	11.08	Aloqaily et al. [205]
36	11.08	Fang et al. [167]
37	10.94	Ahad et al. [206]
38	10.88	Deng et al. [207]
39	10.85	Singh et al. [144]
40	10.20	Nica et al. [208]

) were read and used for the future research agenda. In contrast to the broader bibliometric dataset, we applied a manual content-based filtering step at this stage to ensure that the selected articles are substantively positioned at the core of the research field. This step was taken to guarantee that the identified future research directions are grounded in publications that meaningfully contribute to the intersection of AI and smart cities.

Table 9. Exemplary future research questions.

Research Area
Smart energy and smart grids	How can the development of big data in smart grids be standardized to facilitate high quality datasets for the training of AI-models? How can the communication infrastructure of smart grids and communication protocols of smart grid devices be standardized? What are the most efficient AI or ML approaches for improving the performance of smart grids? How can different energy sources, such as renewable energy from wind, water, and biogas, be integrated into an efficient AI-based energy management framework for smart cities?
Smart citizens and community engagement	How can AI be used to leverage behavioral insights to encourage environmentally friendly behavior? How can the acceptance of citizens towards AI-based solutions be guaranteed and measured? What potential do social media data offer for AI to get insights into citizens and citizens’ concerns and wishes? Which data and which social networks are most suitable for obtaining information about citizens, and what are the specific features and challenges in each case? What is the potential of theories such as UTAUT and TAM to offer additional insights into the acceptance of AI solutions in smart cities?
Integration with other technologies	What would be an effective framework for standardized blockchain testing within smart cities? How can blockchain technology and AI be integrated to address issues of sustainability, scalability, security, and privacy? How can data in smart cities best be stored and how can efficiency and safety be balanced in the most appropriate way? How can blockchain be used within smart cities to enhance AI solutions? How can AI technologies be used to assist within blockchain use-cases?
Disaster management	How can smart cities use AI technologies to predict the occurrence and spread of forest and wildfires, and what AI tools are most suitable for this purpose? How can smart cities use AI technologies to predict the occurrence and spread of floodwaters, and which AI tools are most suitable for this purpose? How can smart cities use and analyze seismic data to detect patterns that may precede earthquakes and how should an early warning system look like? How can smart cities use AI effectively to assess damage to buildings and infrastructure following natural disasters? How can AI predict droughts by analyzing climate and water usage data, and how can it assist in managing water resources more efficiently? How can different cities cooperate to share data, learnings, and insights to optimize the applicability of AI to fight natural disasters?
UAV	How can the privacy and security in UAV’s communication networks be improved? How can the involvement of drones in SCs be regulated, while addressing privacy and security issues? For blockchain-assisted UAV networks, how can off-chain storage be integrated to solve the issue of efficiently storing various types of data, that are too large to store on the blockchain? For blockchain-assisted UAV networks, how can the issue of scalability be solved to ensure QoS? How can security, reliability, and trustworthiness be established within AI in UAV? What are the best path planning approaches for disaster management with UAV?

shows a summary of exemplary future research questions.

5.1. Smart Energy and Smart Grids

The smart grid infrastructure and applications allow cities to analyze and manage energy usage from a single home to an entire city. This enables cities to better predict peaks in energy consumption, account for possible shortages, maintain load balance, optimize costs, set priorities, make decisions about the distribution and acquisition of energy, and plan urban projects with foresight [209]. However, as [15] point out, there is a lack of standardization regarding the development of big data in smart grids. In this context, it is also important to standardize the smart grid’s communication infrastructures as well as the communication protocols used by its IoT devices. Furthermore, ref. [15] explain that machine learning (ML) is key to managing communication technologies and energy supplies. Therefore, future research should also focus on identifying the best AI approaches to optimize the smart grid’s performance. In this regard, deep reinforcement learning (DRL) techniques could be explored for real-time energy demand forecasting and adaptive grid load balancing, allowing smart grids to dynamically adjust to fluctuations in consumption. Additionally, federated learning could be investigated to enhance distributed energy management while ensuring privacy-preserving data sharing across different grid entities. Furthermore, ref. [210] note that future research should focus on the evaluation of different AI models to improve the smart city’s (SC’s) energy management, specifically in predicting electrical energy consumption in summer and winter periods as well as peak demands. Also, ref. [211] suggest that future research should focus on further analyzing the behavioral characteristics of entities using smart grids to improve short-term electrical load forecasting. Finally, it might be beneficial to create an AI-based energy management framework for smart cities, which integrates different types of energy sources, such as renewable energy from water, wind, or biogas [210]. In this regard, hybrid AI models combining neural networks with optimization algorithms (e.g., genetic algorithms or particle swarm optimization) could enhance decision-making for renewable energy allocation, grid stability, and demand response strategies.

5.2. Integration with Other Technologies

From the beginning, the smart city vision was built by integrating advances in emerging technologies, such as IoT and AI. As shown above, one of the most promising technologies with many smart city use cases is blockchain technology due to its transparent, immutable, and secure characteristics. As shown by [141], blockchain and AI technologies can also be combined within smart cities. Such a combination can manifest in various ways (for a recent overview, see [139]). First, blockchain can be used to help overcome problems and shortcomings associated with AI technologies, such as blockchain-based management of AI data (see, e.g., [212,213]) or improving the explainability of AI systems [214]. Additionally, AI can be used to enhance blockchain solutions, for instance by enhancing the mining process [215] or supporting smart contracts [216]. However, the combination of AI and blockchain is still in its infancy and thus has several open research avenues, both in general and applied to smart cities. First, many developers are suspected of falsifying their blockchain performance. Therefore, there is a need for an effective framework for standardized blockchain testing [141]. Furthermore, especially in cities where certain regulatory requirements might be necessary, the question arises as to which blockchain technology should be used. Although organizations such as IEEE, ITU, and NIST are working on blockchain standards, this area requires further research [139]. Furthermore, when cities aim to use blockchain networks and AI in combination, the question of how the blockchain system should be governed and maintained [139] becomes crucial. Aside from blockchain, digital twins are another technical development that is widely investigated in AI in smart cities. Digital twins are virtual representations of physical entities that, combined with AI, can reveal detailed insights into the simulated entity [217]. However, despite their potential, digital twins require a lot of data, just like AI systems. Therefore, ref. [207] emphasize that the rapid surge in data generated by smart city devices has made managing and storing this information in data centers both inefficient and unreliable. Uploading data to cloud servers can lead to significant delays and potentially compromise data integrity. As a result, it is essential to find a balance between efficiency and security and address the question of how the increasing amount of data generated within cities can best be stored.

5.3. Smart Citizens and Community Engagement

Considering the needs and preferences of citizens is crucial for the successful implementation of AI-based solutions in smart cities. Given that, ref. [176] emphasize that it may be valuable to explore methods for involving local communities, stakeholders, and citizens in the design, implementation, and monitoring of AI and IoT solutions for smarter eco-cities, since the active participation of these stakeholder groups can result in more inclusive and effective outcomes. One potential approach is to utilize explainable AI (XAI) to enhance transparency and trust in AI-driven decision-making, allowing citizens to better understand how and why AI systems make certain recommendations.

Additionally, social media data might offer valuable insights that can help cities to better understand urban processes and citizen acceptance. For instance, ref. [218] developed a deep learning-based framework using social media data to assess urban park management across various factors, applying it to seven parks in Wuhan to evaluate their performance and identify management issues, while [219] applied AI tools to analyze 114,390 Reddit comments about Saskatoon to explore citizen engagement, using topic modeling and sentiment analysis to identify twelve key themes related to smart city governance. Future research should investigate which social media platforms provide the most relevant insights for urban planning and how natural language processing (NLP) techniques—such as transformer-based models like BERT or GPT—can be leveraged for real-time sentiment analysis and public opinion tracking regarding smart city initiatives. Given the need for sustainability in smart cities, it might also be worth examining how AI can utilize behavioral insights to promote eco-friendly actions among citizens, including energy conservation and waste reduction [176]. Reinforcement learning could be explored to design personalized incentives for sustainable behaviors, while affective computing (AI systems that interpret human emotions) might help tailor eco-friendly messaging strategies to different demographic groups. To achieve this, researchers might consider technology acceptance models, like the Unified Theory of Acceptance and Use of Technology (UTAUT) [220], as these may provide further insights into the acceptance of certain AI solutions within smart cities.

5.4. Disaster Management

Natural disasters can lead to catastrophic consequences for smart cities, affecting both human life and critical infrastructure. For instance, urbanization and climate change affect the urban hydrological cycle, contributing to severe water stress and urban flooding, which in turn impacts the performance and growth of smart cities. To address these challenges, sustainable and integrated urban flood management strategies are essential in the planning and development of smart cities [221]. Furthermore, wildfires are increasingly endangering people and urban areas, and, as temperatures continue to increase, wildland fires are anticipated to become more frequent and severe, posing significant threats to human settlements, ecosystems, and the environment [222,223,224]. AI can play a crucial role in predicting risks and natural disasters by enabling real-time monitoring, early warning systems, and decision support. Deep learning models, such as convolutional neural networks (CNNs) and long short-term memory (LSTM) networks, can analyze satellite imagery and sensor data to detect early signs of wildfires, floods, and other disasters. For instance, CNNs can be trained to recognize smoke and fire patterns, while LSTMs can process historical weather and climate data to improve the accuracy of extreme event forecasts. The potential applications are not limited to water stress and wildfires but also extend to earthquakes, tsunamis, and hurricanes. Additionally, future research should investigate how different cities can cooperate by sharing data, lessons learned, and insights to optimize the use of AI in combating natural disasters.

5.5. UAVs

The advances in AI have allowed researchers to develop and operate smart autonomous systems. Unmanned aerial vehicles (UAVs)—or drones—are one example of such a system with many smart city (SC) use cases, such as their use by the police for emergency situations or in smart agriculture to monitor and manipulate crops crop [225]. However, with the increasing deployment of drones, it is important to improve privacy and security in UAV communication networks [226]. Therefore, it needs to be determined how drones’ involvement in SCs can be regulated and standardized [225]. Furthermore, drones’ resources need to be optimized to process large volumes of sensor data, provide edge servers, or transfer updated versions of learning models [225]. This is especially important for applications of UAV networks that are blockchain-assisted. In that case, the issue of blockchain scalability also needs to be addressed to ensure high levels of service quality. Lastly, one of the main use cases of UAV networks in SCs is disaster management. Therefore, future research should focus on identifying and optimizing the best approaches for path planning of drones, as well as for pre-disaster preparedness and response planning [225]. Furthermore, as [201] point out, several open research questions remain regarding UAVs, especially concerning security, reliability, and the trustworthiness of AI in UAV operations and applications.

6. Discussion and Conclusions

In this study, we aimed to investigate the application of AI technologies within smart cities. To do this, we posed and answered three research questions. First, we examined the present status and structure of knowledge in this domain, identifying how AI is currently applied in smart cities. Second, we explored the major themes that have been extensively studied in the existing scientific literature, providing insight into the areas that have garnered significant research attention. Lastly, we identified potential avenues and emerging themes for future research, highlighting gaps and opportunities that can guide further exploration in the application of AI for urban use cases. We conducted a bibliometric study and applied performance analysis to identify the most influential countries, disciplines, and affiliations, as well as the most relevant funding institutions and outlets. Furthermore, we used keyword co-occurrence to map the knowledge structure and to identify thematic clusters and research topics. Based on this, we were able to identify five thematic areas: complementary technologies and security, intelligent transportation and smart mobility, AI-based energy efficiency, computer vision and object detection, and governance and urban planning. Additionally, based on the findings of the bibliometric study and an analysis of the 40 most relevant papers (based on their normalized total citations (NTC)) in our sample published between 2020 and 2024, we derived future research questions for four different areas: smart energy and smart grids, integration with other technologies, smart citizens and community engagement, disaster management, and UAVs.

As such, this study makes a distinctive contribution by providing the first comprehensive examination of AI in smart cities through a bibliometric study that combines a structured literature analysis. Furthermore, while previous reviews have primarily focused on specific subfields. Examples include AI for waste management [167], safety and security [22], or the convergence of AI and blockchain within smart cities [227]. In comparison, our study adopts a holistic perspective on AI in smart cities as a whole. mapping the intellectual structure of the field and identifying key research components, thematic trends, and interdisciplinary connections. By leveraging bibliometric techniques, we offer an objective, data-driven assessment of the evolution and current state of research, while our complementary literature analysis ensures a deeper contextual understanding. With this combined approach, we aim to extend beyond existing work but also provides valuable insights for future research directions and practical applications in the development of AI-driven smart cities.

Our study has several implications for both practitioners and scholars. First, for researchers, our study is beneficial as it shows the evolution of research trends in the domain of AI in smart cities as a whole, enabling scholars to see how the field has developed over time. By highlighting which topics have dominated the conversation and which areas remain underexplored, the study offers researchers an overview of where the intellectual focus has been and where new contributions could be made. Together with the future research avenues we outlined, this might help guide future research endeavors. Moreover, the study’s identification of emerging trends helps researchers position their work at the cutting edge of the field. By understanding which themes are gaining traction, scholars can tailor their research to align with the latest developments, ensuring their work remains relevant and impactful. At the same time, the study underscores the multidisciplinary nature of AI in smart cities, which suggests that future breakthroughs will likely come from interdisciplinary collaboration. Researchers from fields such as computer science, urban planning, and environmental studies should therefore find common ground, working together to address the complex challenges that smart cities face.

Next to that, the findings can be used by practitioners, such as organizations, managers, cities, and industry-based researchers, to further develop the current body of knowledge on AI in smart cities and advance its adoption and implementation. Furthermore, practitioners can use the most influential identified articles to discuss the design, benefits, and issues that may affect the implementation of AI in smart cities. Next, practitioners who would like to drive technological advancements in their organizations, institutions, or cities may use the findings of this study to understand the scope and boundaries of AI’s applicability in improving applications in smart cities across different sectors. Finally, practitioners may benefit from the findings of the bibliometric analysis and future research agenda to understand the boundaries of research within individual sub-domains of interest and use the results as a foundation for new implementations in specific domains, such as smart energy and smart mobility.

It is important to note, however, that there are also disadvantages and challenges associated with the use of AI in smart cities, particularly concerning privacy, bias, and ethical considerations. The vast amounts of data collected from sensors, cameras, and digital interactions raise significant privacy concerns, as unauthorized access or misuse of personal information could undermine public trust. Additionally, AI models can inherit biases from training datasets, potentially leading to unfair treatment in decision-making processes related to public services, resource allocation, and security. To address this, researchers and practitioners should prioritize bias detection algorithms, explainable XAI, and the use of diverse and representative datasets. Furthermore, the increasing reliance on AI-driven governance introduces ethical questions regarding accountability and transparency. AI-powered decision-making in urban management—such as law enforcement, traffic control, and energy distribution—requires clear oversight mechanisms. Implementing AI ethics frameworks, transparent governance policies, and human-in-the-loop systems can help ensure that AI applications align with societal values and democratic principles. By proactively addressing these challenges, cities can harness AI’s full potential while maintaining public trust and ensuring inclusive and responsible smart city development.

Finally, our study and the applied method itself are subject to certain limitations. First, we relied solely on a single database, namely Scopus, for collecting the bibliometric data. While Scopus is a comprehensive and widely respected source that encompasses numerous journals and conferences, this choice inherently restricts the scope of outlets covered. Additionally, since our goal was to investigate the current state of research on AI within smart cities, our sample was confined to journal articles, conference papers, and reviews. Non-scientific publications, such as grey literature and practitioner reports, were intentionally excluded. This limitation also extends to non-English and non-final articles, which were not part of our initial data collection. As a result, our study focuses strictly on scientific literature, reflecting the current state of research without accounting for practical developments. It is also important to acknowledge that bibliometric studies cannot be seen as a substitute for qualitative analyses. While bibliometric methods are useful for quantitatively evaluating large bodies of literature, qualitative approaches are necessary for a more in-depth exploration of specific aspects of smart city research.