Artificial Intelligence for Smart Cities: A Comprehensive Review Across Six Pillars and Global Case Studies

Abstract

Rapid urbanization in the twenty-first century has significantly accelerated the adoption of artificial intelligence (AI) technologies to address growing challenges in governance, mobility, energy, and urban security. This paper explores how AI is transforming smart city infrastructure, analyzing more than 92 academic publications published between 2012 and 2024. Key AI applications ranging from predictive analytics in e-governance to machine learning models in renewable energy management and autonomous mobility systems are synthesized domain-wise throughout this study. This paper highlights the benefits of AI-enabled decision making, finds current implementation barriers, and discusses the associated ethical implications. Furthermore, it presents a research agenda that stresses data interoperability, transparency, and human–AI collaboration to steer upcoming advancements in smart urban ecosystems.

1. Introduction

Managing city infrastructure, resources, and services has become much harder because cities have grown quickly in the last few decades. Traffic jams, pollution, energy use, and public safety are just a few more essential problems as cities grow. The idea of “smart cities” has come up to discuss these problems. It uses cutting-edge technology to make city life better. Artificial Intelligence (AI) [1,2,3] is at the heart of this change because it offers new ways to make cities more sustainable [4,5]. Combining AI with city planning and management is vital for reaching sustainable development goals. Al-Raeei [4] talks about how AI could help cities grow in a way that is good for the environment by using smart systems to make the best use of resources and provide services. Sharma [5] talks about how AI affects the design of future cities by making infrastructure better and getting people more involved. A bibliometric analysis by Karger et al. [6] shows increasing research interest regarding how AI can be used in cities. This suggests that city management is moving toward being more data-driven. Ullah et al. [7] give a complete picture of how AI and machine learning are used in smart cities, focusing on how they help with traffic management, energy efficiency, and public safety.

AI also has a significant effect on the environment of smart cities. Gavade [8] looks at digital transformation strategies that make the environment more sustainable. Yigitcanlar et al. [9] introduce the idea of “Green AI,” which focuses on making AI systems that use less energy. Ortega-Fernández et al. [10] talk about how AI can help cities become more innovative and sustainable. AI must work with technologies like the Internet of Things (IoT) and big data analytics for smart city development. Yao [11] looks at how these technologies can be used in cities as a whole, and Van Hoang et al. [12] looks at how they can work together to change cities into smart cities. Bibri (2021) [13] suggests a data-driven method for strategic urban planning using AI to make better decisions. AI-powered predictive analytics can make cities much more resilient. Garcia [14] talks about how AI-powered predictive models can help cities deal with problems before they happen. Rakshit talks about cloud-based AI solutions [15] that make data analysis in smart cities more efficient, allowing cities to respond to real-time changes. Bibri [16] goes into more detail about the ideas behind urban intelligence functions and how to use AI in different city systems.

AI helps make city designs more efficient and long-lasting when planning cities. Hadiyana and Ji-hoon [17] look at AI-driven ways to plan cities, and Piri [18] talks about how adding smart infrastructure can make cities more resilient. Puliafito et al. [19] look at smart cities as cyber-physical systems and discuss the problems and technologies that make them possible. Smart cities also need to be safe and connected. Lv et al. [20] look into how computational intelligence can help protect digital twins and big data in cyber-physical systems. Sheraz et al. [21] give a complete overview of AI-enabled digital twin networks in 6G connectivity and point out where smart city infrastructure should go. Smart cities are changing how they are run in reaction to the effects of AI. Bokhari and Myeong talk about how AI can help with knowledge management and e-service delivery, but e-governance frameworks can change how it works [22]. Al-Mushayt [23] looks at how AI can automate e-government services to make them more efficient and easier to access. Bokhari and Myeong look at AI’s effects on e-governance and cybersecurity from the points of view of different stakeholders.

For smart cities to work, they need to focus on the needs of their citizens. Obedait et al. [24] argue for smart government services to be delivered with a focus on citizens, ensuring that new technologies meet the needs of the public. Voelz et al. [25] stress the importance of making services that people can use that put citizen engagement and satisfaction first. Raghav et al. [26] look into the moral and legal ways to use AI and robotics in smart city governance to ensure that innovation is done responsibly. Big data and new technologies are crucial for monitoring and growing smart cities. Ma et al. [27] talk about how big data can be used to track how cities are growing. George and George [28], on the other hand, look at how new technologies can be used together for social innovation in the context of Society 5.0. Kasinathan et al. [29] look at how disruptive technologies, such as Industry 5.0 and smart cities, can help us reach our sustainable development goals. Schwarz-Herion (2019) [30] discusses how smart cities can help achieve sustainable development goals. He stresses the importance of ensuring that new technologies align with global goals. Last but not least, smart cities need to have sustainable mobility. Anthony Jr. [31] gives an overview and a conceptual model for sustainable mobility governance, showing how to make sound urban policy.

Although there is a growing body of literature on the use of AI in smart cities, many existing reviews exhibit significant limitations. First, much of the research adopts a domain-specific focus, examining areas such as mobility or governance in isolation rather than analyzing the interconnections among urban systems. Second, most studies are connected only to specific regional contexts, which limits their broader global relevance. Third, essential themes such as ethical governance, citizen participation, and digital inequality are often underexplored or entirely omitted. This review addresses these gaps by showing a comprehensive six-pillar framework that analyzes AI’s role across interconnected urban domains, namely governance, economy, mobility, environment, living, and people. Via a synthesis of bibliometric analysis, international case studies, and cross-sectoral insights, this paper provides a structured yet critical assessment of AI’s transformative impact on smart cities. It recognizes global best practices, highlights common barriers, and underscores the need for comprehensive, ethical, and strong AI integration in future urban systems. (Figure 1) offers a conceptual overview of our main contributions on how artificial intelligence plays a crucial role in urban transformation. The main contributions of this research work are listed below.

• This paper gives a full look at how AI is used in six important areas of smart cities: governance, economy, mobility, environment, living, and people. This thematic framework gives a big-picture view of how AI technologies are changing cities by making services better, managing resources better, and getting citizens more involved.
• It has in-depth case studies from major cities like Singapore, Estonia, Amsterdam, and Delhi that compare different ways to use AI. These examples show important success factors, like a clear policy vision, partnerships between the public and private sectors, and design that puts citizens first. They also show problems, like planning that is not coordinated and digital inequality.
• The analysis of AI’s role in making cities more resilient to climate change is a significant part of the work. This paper discusses how AI models can be used to predict floods, monitor air and water quality, save energy, and reduce the effects of urban heat islands. It connects smart technologies with sustainability and disaster risk reduction.
• This paper stresses the importance of using AI responsibly by making it clear, open, and open to participation. It talks about important issues like algorithmic bias, surveillance, and privacy, and it calls for ethical frameworks and citizen participation in smart city AI systems.
• This paper discusses new technologies like Edge AI, 5G, federated learning, and blockchain integration. It calls for research and policy frameworks that cross disciplines and focus on resilience, equity, and long-term urban sustainability. This will make sure that AI is used for city development that is open to everyone and can change as needed.

2. Methodology

This review adopts a structured conceptual review methodology that combines thematic content with bibliometric mapping to examine how Artificial Intelligence (AI) is implemented across the six foundational domains of smart cities. The goal is to synthesize trends, identify implementation challenges, and extract cross-regional insights from diverse governance and technology contexts.

2.1. Theoretical and Conceptual Framework

Rather than using a single theoretical model, this review synthesizes recurring conceptual patterns found across the reviewed literature to frame the role of Artificial Intelligence in smart city transformation. The analysis draws from existing studies that explore the interplay between AI deployment, institutional governance, and citizen engagement. Several reviewed papers shows that AI in smart cities cannot be seen as a purely technical solution; it is embedded within broader institutional, cultural, and political structures. For example, papers on Estonia’s e-Governance and Singapore’s multilingual virtual assistants shows how administrative maturity and policy continuity affect the success of AI adoption [32,33]. Similarly, studies on Toronto’s smart health clinics and Amsterdam’s energy grids stress the importance of transparency, trust, and participatory design in building citizen confidence [34,35]. This method of conceptual framing treats AI-enabled urban systems as socio-technical arrangements—where algorithms, infrastructures, policies, and people co-evolve. Rather than applying external theory, this review finds theoretical contributions emerging from real-world implementations: highlighting themes of digital inclusion, algorithmic transparency, ethical foresight, and governance responsiveness [32,36]. These themes guide our comparative analysis across case studies and inform the synthesis of barriers, outcomes, and best practices.

2.2. Databases and Search Strategy

To ensure both depth and relevance, sources were retrieved from six scholarly databases: IEEE Xplore, ScienceDirect, SpringerLink, MDPI, Scopus, and Google Scholar. The time frame for literature collection was restricted to the period between 2015 and 2024 to capture both foundational work and the most recent developments in the field. Search terms combined general keywords such as “smart cities,” “AI in urban planning,” and “IoT in cities,” with more domain-specific terms like “mobility optimization,” “AI-based governance,” and “AI infrastructure planning.”

2.3. Inclusion and Exclusion Criteria

The literature selection followed well-defined inclusion and exclusion criteria. A purposive sampling approach, appropriate for conceptual mapping, was adopted to ensure diversity in geography, income context, and governance models. Studies were included if they were peer-reviewed, written in English, and focused on the application of AI—such as machine learning algorithms or digital twin technologies within one or more of the six smart city pillars: governance, economy, mobility, environment, living, or people. Furthermore, preference was given to empirical or conceptual studies that reported measurable outcomes such as improved efficiency, accessibility, or service responsiveness. Conversely, studies were excluded if they focused solely on rural or non-urban environments, lacked a clear methodological framework, or presented opinion-based commentary without empirical grounding or outcome metrics. Editorials, unreviewed articles, and speculative essays were also excluded to maintain academic rigor.

2.4. Analytical Framework

Each selected source was thematically classified according to its smart city domain and then analyzed across three comparative dimensions. The first dimension focused on implementation metrics, such as energy efficiency improvements, congestion reduction, or levels of user engagement. The second assessed the underlying technological methods, including the specific types of AI models employed and system design. The third dimension evaluated contextual enablers, such as the presence of supportive policy frameworks, institutional maturity, and infrastructure readiness. Case studies were deliberately selected to reflect a diverse range of geographic regions and income levels, enabling cross-comparative insights across cities with varying technological capacity and governance types. In total, approximately 90 scholarly sources were shortlisted and reviewed. These were synthesized to identify prevailing challenges, underlying conditions, and replicable best practices in AI-driven smart city initiatives. The analysis process was guided by theoretical perspectives including socio-technical systems theory, responsible innovation, and algorithmic governance. These lenses provided a foundation for interpreting how AI interacts with policy structures, citizen participation, and institutional maturity across diverse urban contexts and power dynamics.

2.5. Bibliometric Analysis of Urban AI Literature Trends

To complement the thematic synthesis, a bibliometric analysis was conducted to map research trends in Urban AI literature from 2012 to 2024. This approach helped to analyze the evolving intellectual structure of the field. This was in addition to the thematic analysis of AI integration in smart city domains. We used keyword co-occurrence data from more than 92 peer-reviewed articles and made a network graph, which is shown in (Figure 2). The bibliometric network visualization was conducted using VOSviewer, a widely adopted tool for mapping keyword co-occurrence, citation clusters, and thematic evolution in scientific literature. VOSviewer(version 1.6.20) was selected for its ability to handle large datasets and generate interpretable cluster-based maps from Scopus and Web of Science data. In this network, nodes are keywords that frequently occur, and the size of each node shows how often the term comes up. The thickness of the edges shows how strongly the terms co-occur. A modularity-based algorithm was used to find color-coded clusters that show different thematic areas. The analysis shows five main research streams. One focuses on algorithmic methods and traffic intelligence, with high-frequency terms like “network,” “machine learning,” and “traffic congestion” showing a strong focus on mobility optimization and predictive modeling. Another focuses on governance and digital policy, linking terms like “governance,” “innovation,” and “digital economy,” which suggests a shift toward institutional applications of AI. A third stream deals with sustainability and Industry 4.0, with recurring ideas like “smart home,” “agriculture,” and “sustainable city.” A fourth group highlights speculative themes like “metaverse” and “platformization,” which reflect futuristic views on AI urbanism. Finally, a cluster built around terms like “carbon emission” and “digital transformation” shows how AI is becoming more closely linked to economic modernization and environmental goals. These bibliometric insights support the conceptual review by highlighting how interdisciplinary and socio-technical themes have emerged in Urban AI discourse.

2.6. Practical Tools for AI-Enabled Smart Cities

Smart cities are using increasingly practical tools and platforms to help AI work in different areas of the city. This study used VOSviewer as a bibliometric visualization tool to create maps of networks of keywords that appear together. Its modularity-based clustering algorithms helped to find the main research topics and areas where AI-driven smart cities were coming together. The SAIoT-SL (Secure AIoT-enabled Smart Living) protocol shows how lightweight authentication and key agreement can be used between edge devices, user interfaces, and cloud layers for AI-enabled home security and automation. The system is protected against common attacks like impersonation and replay, and it has been tested with formal security verification tools like Scyther.

Explainable AI (XAI) methods like SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model-agnostic Explanations) are being used to make sense of predictions made by black-box AI models like LSTM, GRU, and random forest. This is especially true for tasks like predicting traffic jams or making predictions about public policy. These tools for understanding are important for building trust in algorithms and holding institutions accountable. Digital twin systems used in smart living environments also use real-time IoT sensor data to mimic systems in the real world. Trained neural networks and fuzzy control mechanisms make it possible for these platforms to make decisions automatically in HVAC systems, security management, and health monitoring. They can also be used to test scenarios, find anomalies, and change the system without needing help from people.

Finally, city-wide platforms like Mobility-as-a-Service (MaaS) ecosystems, such as Helsinki’s Whim, are examples of advanced AI-driven infrastructure that brings together different types of transport into one app. To make sure that travel planning is reliable and can be done in more than one way, these services use blockchain-based identity layers, predictive routing algorithms, and real-time IoT feeds. These practical tools bring together AI theory and real-life city functions, showing how data, intelligence, and citizen experience will all work together in next-generation smart cities.

3. Pillars of Smart Cities and AI Applications

This section builds upon the methodological framing by examining how AI technologies are applied across six foundational domains—or “pillars”—of smart cities: governance, economy, mobility, environment, living, and people. Each pillar is presented with key technologies, global examples, and contextual challenges to show the breadth and variation of AI use cases. Figure 3 provides a conceptual snapshot of these thematic domains.

3.1. Smart Governance

Smart governance is the strategic application of emergent information and communication technologies (ICT) to improve public decision-making, encourage citizen engagement, and strengthen stakeholder collaboration. The literature indicates that smart government forms the foundation for developing smart governance through both traditional ICT tools (e.g., e-participation portals and open data initiatives) and advanced AI systems capable of adaptive learning, predictive analytics, and real-time decision support to promote transparency, information sharing, and citizen participation [37]. Smart governance is essential in smart city endeavors involving complex interactions among governments, citizens, and private organizations. Pereira et al. [37] suggest a framework for building new governance models capable of addressing challenges in the digital society, emphasizing openness, collaborative governance, and informed decision-making. A practical study by Lopes [38], based on interviews with participants from Brazil, Singapore, Colombia, Portugal, and Uruguay, reveals that smart cities and e-government have evolved along parallel trajectories, converging towards smart governance. The study asserts that innovation, advanced technologies, and smart governance are prerequisites for creating sustainable, creative, and resilient cities.

The increasing integration of artificial intelligence introduces new dimensions to smart governance. Rajagopal et al. [39] highlight that as AI systems become pervasive across sectors, public governance must develop appropriate frameworks to align AI development and deployment with ethical, legal, and societal standards. Their study suggests a complete AI governance model that includes regulatory, ethical, and operational elements to help smart government grow. AI governance frameworks are necessary for dealing with new risks, making sure that innovation is done responsibly, and building public trust in public administration. Smart governance has a real-world effect, as shown by these examples. Estonia’s e-Residency program, which will receive more security features in 2023, gives people from all over the world digital identities that let them use Estonian e-services, sign documents digitally, and register a business without being there in person. This project sets a standard for digital governance across borders around the world. Singapore’s Smart Nation initiative connects AI-powered citizen service platforms across more than 70 government agencies. These platforms offer real-time information, personalized responses, and support in multiple languages.

As governments increasingly rely on AI, they need not only technical effectiveness but also strong ethical protections. As predictive systems affect public policy, how resources are used, and how people interact with each other, the risks of algorithmic bias, too much surveillance, and unclear decision-making grow. To stop discrimination and protect civil liberties, governance frameworks must include explainable AI (XAI), transparency audits, and algorithmic accountability boards. Governments must also make sure that they follow global privacy standards like the GDPR by limiting the amount of data they collect and only allowing people to access it with their permission. These worries are even more important when AI tools move beyond simple tasks like providing information and into areas like policing, welfare distribution, or determining who is eligible. As of 2024, Singapore is adding AI chatbots that can learn and adapt to these systems, and they are also making them more useful in healthcare, transportation, and city services. Vadisetty [40] also shows how AI could make government work better by using a case study of traffic management. An AI framework was made to predict traffic jams over a long period of time by using machine learning models like Multilayer Perceptron (MLP), Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), LightGBM regression, and random forest regression. The MLP model was proven to be better based on performance tests that used metrics like Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and $R^2$ score. We used explainable AI methods to help us understand the results of the models, which helped build trust, accountability, and openness in AI-based governance systems.

Table 1. Smart governance models and technologies.

Governance Model	Technologies	Examples	Benefits	Challenges
Digital identity	AI, blockchain	Estonia e-Residency	Transparency, efficiency	Privacy, security
AI-driven citizen services	AI chatbots	Singapore’s Ask Jamie	Real-time responses	Algorithm transparency
traffic management	MLP, LSTM, GRU	Vadisetty model	Congestion prediction	Data accuracy

summarizes representative smart governance models, the AI technologies powering them, their demonstrated benefits, and the ongoing implementation challenges observed across leading case studies.

3.2. Smart Economy

The smart economy represents a critical pillar of smart cities, integrating advanced technologies such as artificial intelligence (AI), big data analytics, IoT, and automation to drive sustainable economic growth, enhance operational efficiency, and promote innovation. Human–machine collaboration is central to the smart economy, where AI and automation synergize with human expertise to transform industries, foster net job creation, and provide strategic advantages for international organizations [41]. AI applications in smart cities span multiple domains, notably smart mobility, smart environment, smart governance, smart living, smart economy, and smart people, with emerging research emphasizing the identification of solutions, barriers, and future directions for AI-driven transformations [42].

In the context of smart infrastructure, AI-driven smart buildings present an important economic and social opportunity. Through the Integrated AI-Driven Smart Buildings Framework (IAI-DSBF), operational efficiency, community engagement, sustainability, and resilience have been substantially enhanced, with empirical results showing a 99.1% increase in building performance, a 98.5% rise in community engagement, and a 99.12% boost in economic growth [43]. These smart buildings integrate reinforcement learning algorithms and rule-based control systems trained on real-time occupancy, energy usage, and environmental sensor data. Automation in these systems is mainly closed-loop, enabling dynamic HVAC control, lighting optimization, and predictive maintenance without continuous human intervention. Such integrated frameworks address challenges like data privacy, regulatory compliance, and equitable access to smart technologies.

At a broader industrial level, the smart economy is deeply intertwined with Industry 4.0 principles, leveraging AI, IoT, and big data to optimize urban planning, agriculture, energy management, and cybersecurity. A systematic literature analysis confirms that smart technologies have led to tangible benefits: smart lighting systems have reduced energy use by 40%, AI-IoT-enabled urban planning has decreased traffic congestion by 30%, and precision agriculture has improved crop yields by 25% while reducing water consumption by 30% [44]. Nevertheless, concerns around data privacy and digital inequalities must be addressed to realize the full potential of the smart economy. Furthermore, in the energy sector, intelligent systems incorporating AI and low-carbon economy models have been crucial for sustainable development, particularly in contexts such as China’s smart energy industry. Optimizations using adaptive particle swarm algorithms on load, price, and climate data have achieved high forecasting accuracy, with median errors of less than 1%, thus supporting efficient scheduling and resource management [45]. Real-world implementations of the smart economy can be seen in Barcelona’s digital startup ecosystem. Barcelona has established itself as a leading innovation hub through initiatives like the 22@Barcelona district, which has fostered thousands of startups in AI, green technology, and urban digital services. As of 2023, the city introduced additional public funding programs and accelerator platforms focused on AI-driven entrepreneurship, green innovation, and smart manufacturing, reinforcing its position as one of Europe’s most dynamic smart city economies. Collectively, these advancements position the smart economy as a catalyst for achieving the Sustainable Development Goals (SDGs) while promoting resilient, low-carbon urban futures.

Table 2. AI in smart economy initiatives.

Economic Domain	Core Technologies	Implementations	Impact	Barriers
Smart buildings	IoT, automation	AI-integrated framework	Efficiency (99.1%)	Privacy
Industry 4.0	IoT, big data, AI	Barcelona 22@	GDP growth (EUR 3B/year)	Digital inequalities
Energy management	AI forecasting	Smart grids (China)	Load mgmt (99%)	Regulatory

summarizes key AI-enabled smart economy initiatives across domains such as smart buildings, energy management, and industrial innovation. It highlights the technologies used, measurable impacts achieved, and ongoing challenges related to privacy, inequality, and regulation.

3.3. Smart Mobility

Smart mobility has emerged as a fundamental component of smart city development, aiming to address urbanization challenges by promoting sustainable, efficient, and inclusive transportation solutions. It integrates advanced technologies such as the IoT, AI, blockchain, and big data to transform the logistics of moving people and goods while reducing ecological footprints [41]. Key features of smart mobility include traffic congestion reduction, route optimization, active and inclusive mobility promotion, and citizen engagement. Mobility-as-a-Service (MaaS), autonomous vehicles, and logistics optimization are among the primary services driving the evolution of smart urban environments [41]. Artificial intelligence plays a pivotal role in reshaping urban transport systems. AI’s deep learning capabilities enable the development of intelligent transportation systems and autonomous mobility solutions that enhance urban livability and resource efficiency [42]. These systems use a mix of deep learning models such as convolutional neural networks (CNNs) for image-based navigation and recurrent neural networks (RNNs) for traffic sequence forecasting trained on mobility logs, GPS trajectories, vehicular sensor feeds, and real-time surveillance data. Depending on deployment, autonomy ranges from AI-assisted route optimization to fully autonomous vehicle decision loops in closed environments. The integration of Connected and Autonomous Vehicles (CAVs), Personal and Unmanned Aerial Vehicles (PAVs and UAVs), and MaaS frameworks signifies a significant shift towards a machine-driven yet human-centric mobility ecosystem. Trust-building, user satisfaction, and risk mitigation remain essential for the widespread adoption of AI-based mobility solutions [42]. Practical implementations of smart mobility are already observable. London’s Transport for London (TfL) has deployed AI-powered predictive traffic tools capable of analyzing real-time road conditions, congestion patterns, and incident forecasting, thereby optimizing traffic flow and reducing emissions. Similarly, Helsinki’s pioneering Mobility-as-a-Service (MaaS) ecosystem integrates multiple transport modes into a single digital platform, allowing for seamless, on-demand mobility planning, payments, and real-time service updates, fundamentally shifting the traditional paradigms of urban transportation.

In addition to technical transformations, the social and human dimensions of smart mobility are gaining prominence. As reviewed by Dartmann et al. [43], the convergence of AI, machine learning, cloud computing, and communication technologies has triggered a rethinking of transport services and platforms, focusing on automation, citizen needs, and broader societal impacts. Future mobility concepts must address not only technical and algorithmic challenges but also human factors such as trust, safety, and social acceptance. The reproduction of cognitive functions through AI technologies further reinforces smart city evolution, where mobility is a central driver [44]. Innovative intersections between AI, robotics, and transportation systems are facilitating the creation of resilient urban ecosystems. Infrastructure-based AI developments, including IoT-driven physical internet frameworks and broader Industry 4.0 transformations, enhance the adaptability and intelligence of transport systems [44]. Integrated mobility projects leveraging connected and autonomous vehicles and smart communication architectures contribute to a comprehensive vision of urban mobility transformation.

Table 3. AI-based smart mobility solutions.

Solution	Technologies	City Examples	Outcomes	Challenges
MaaS	AI, Blockchain	Helsinki (WIM)	Lower Car Usage (20%)	Integration
Connected vehicles	Edge AI	London AI tool	Reduced congestion (12%)	Risk mitigation
Active mobility	IoT sensors	Helsinki footprint app	Eco-engagement	Digital literacy

synthesizes key AI-powered mobility initiatives, illustrating how cities like Helsinki and London are applying emerging technologies to optimize transportation systems, promote sustainability, and tackle urban mobility challenges.

3.4. Smart Environment

The smart environment paradigm emphasizes the convergence of Artificial Intelligence (AI), Internet of Things (IoT), big data, and edge computing to enhance the design, construction, and maintenance of smart buildings and cities [45]. Recent research focuses on developing AI and machine learning technologies to empower intelligent urban infrastructures while addressing real-world challenges such as cybersecurity, privacy, and sustainability. Emerging topics include AI-driven data analytics, hybrid human–machine computing, and IoT-enhanced smart environment engineering [45]. While IoT and edge computing primarily support sensing and communication, AI models such as time-series forecasters, anomaly detectors, and deep learning classifiers enable automated interpretation, adaptation, and decision-making, which go beyond passive data collection. A significant innovation is the concept of smart radio environments, where reconfigurable intelligent meta-surfaces are embedded into walls and objects, allowing for environmental control over radio wave propagation and reducing the need for new signal emissions [46]. These systems use edge-deployed CNNs, time-series regression models, and anomaly detection algorithms trained on high-frequency sensor data (e.g., temperature, pollution, energy usage) and often function autonomously for threshold-based alerts and optimization routines in buildings and environmental nodes. These meta-surfaces enable seamless, energy-efficient wireless connectivity by recycling existing radio waves, presenting a transformative opportunity for next-generation wireless networks. At a broader level, the convergence of AI, IoT, and big data technologies is driving the realization of environmentally sustainable smart cities [47]. This direction is supported by foundational work such as that of Nazari et al., who highlight the importance of sustainable water reuse strategies in parched and urban regions [48,49] and further explored the importance of long-term planning for freshwater storage and demand in urban infrastructure systems [50].

Addressing challenges related to environmental degradation and climate change, these cities integrate data-driven approaches to align urban growth with the Sustainable Development Goals (SDGs). Amsterdam’s smart energy initiatives, for example, employ AI for real-time energy optimization and flexible grid management, reducing carbon emissions while improving efficiency. Similarly, Delhi’s air quality forecasting systems leverage machine learning models under initiatives like SAFAR to predict PM2.5 and PM10 levels, enabling preemptive public health interventions and environmental policymaking. The SAFAR system, for instance, utilizes ensemble regression models and seasonal decomposition to generate pollutant forecasts from satellite data, traffic flow datasets, and meteorological inputs. Bibliometric analysis reveals that research on environmentally sustainable smart cities has significantly accelerated in the digitalization era, particularly post-COVID-19 [47]. Moreover, explainable and spatially robust AI models have become very important tools. Majnooni et al. created deep learning models that can be understood along with feature importance analysis to keep an eye on the quality of water in reservoirs in a more open way [51]. Zamani et al. showed how a WT–CNN–GRU hybrid model can accurately predict the quality of water in a reservoir when conditions change [52].

Fooladi et al. added to this field by suggesting a clustering-based hybrid framework that includes spatial generalization and uncertainty quantification to make water quality modeling more accurate [53]. At the same time, Karimi et al. used real-time weather forecast data to predict changes in urban surface temperature. This is a useful way to reduce the effects of heat islands in densely populated areas [54]. Their later conceptual framework built on this by combining environmental risk and social vulnerability analysis to improve resilience planning for complicated urban settings [55]. However, despite these benefits, the adoption of these technologies also raises concerns regarding environmental costs, ethical risks, and regulatory complexities that must be carefully managed. In terms of real-time environmental monitoring, platforms like SEMAR (Smart Environmental Monitoring and Analytics in Real-Time) demonstrate how AI-integrated IoT systems can collect, process, and analyze sensor data efficiently [56]. AI techniques such as predictive analytics, object detection, auditory perception, and natural language processing enhance the capabilities of IoT platforms, supporting faster deployments and smarter environmental monitoring solutions. Furthermore, studies on AI applications in smart environments highlight the critical role of machine learning in addressing pressing environmental challenges such as air quality monitoring, water pollution detection, radiation assessment, and sustainable urban planning [57].

Table 4. Smart environment strategies and results.

Strategy	Technologies	Example	Benefits	Risks
Energy management	AI grid optimization	Amsterdam grid	Energy savings (25%)	Privacy
Air quality monitoring	ML analytics	Delhi SAFAR	Health interventions	Infrastructure
Smart radio env.	Intelligent surfaces	Wireless solutions	Efficient signals	Complexity

summarizes key AI-driven strategies in environmental sustainability, including smart grid optimization in Amsterdam, predictive air quality systems in Delhi, and the adoption of intelligent surfaces for more energy-efficient urban communication infrastructure.

3.5. Smart Living

Smart living environments leverage AI, IoT, and emerging technologies to transform residential spaces into intelligent, energy-efficient, and user-centric ecosystems. Integrating sensors and digital twins (DT) provides real-time environmental awareness, while AI models enable predictive personalization, such as adaptive HVAC control, anomaly detection, or behavior-driven lighting, moving beyond rule-based automation to intelligent decision-making [58]. AI-driven simulation frameworks embedded within living spaces allow for dynamic interior control, including adaptive lighting, HVAC optimization, security enhancement, and predictive maintenance, leading to substantial reductions in energy consumption. Digital twin models facilitate real-time monitoring and simulation of smart home environments, enabling proactive decision-making to optimize both functionality and sustainability [58]. These systems implement lightweight neural networks and fuzzy logic controllers trained on real-time home sensor data (e.g., light, motion, energy use), enabling fully autonomous control loops for environment adaptation without human input.

Smart systems in residential housing are increasingly designed for continuous functionality, ensuring seamless interactions across lighting, ventilation, and curtain automation [59]. By incorporating feedback loops and autonomous control strategies, modern smart homes provide users with enhanced convenience, security, and overall satisfaction. Recent innovations in decentralized smart home management leverage local edge computing nodes in addition to cloud services, ensuring lower latency, better resilience, and increased privacy. Securing smart living ecosystems has become paramount with the proliferation of interconnected AIoT devices. The proposed Secure AIoT-Enabled Authenticated Key Agreement Technique for Smart Living Environments (SAIoT-SL) offers resilient security by enabling authenticated communication between cloud servers, user interfaces, and sensors, significantly mitigating risks of interception, replay, and impersonation attacks [60]. Formal security analyses using the Scyther verification tool have validated the robustness of SAIoT-SL protocols. Moreover, comparative studies show that SAIoT-SL achieves lower computational overhead and improved resilience compared to traditional IoT authentication schemes [60]. Recent studies in 2024 further emphasize the necessity of lightweight encryption models and quantum-resilient algorithms for next-generation AIoT-based smart homes, indicating critical directions for future research.

Real-world implementations of smart living concepts are increasingly visible. Toronto’s Smart Health Clinics, integrated within residential smart neighborhoods as of 2023, utilize AI-powered diagnostic systems, telemedicine services, and predictive health analytics embedded into smart homes, providing proactive and personalized healthcare. Simultaneously, Dubai’s AI learning systems, expanded in 2024 under the Smart Dubai 2030 agenda, embed AI-driven educational platforms within residential environments, enabling personalized learning pathways, skill development, and lifelong education directly from smart home interfaces. Furthermore, energy conservation strategies, including AI-driven light control, appliance scheduling, and predictive load balancing, have gained prominence as part of smart living initiatives. New research in 2023 explores integrating federated learning models into smart homes, allowing for localized training of AI models without sharing sensitive user data, thus enhancing privacy while maintaining intelligent system performance. Federated learning here enables on-device training of models for user preference prediction, anomaly detection, and load balancing, reducing dependency on centralized cloud computation while enhancing data privacy.

Table 5. Technologies for smart living.

Solution	Technologies	Example	Outcomes	Challenges
Smart home	Digital twin, AI	Toronto clinics	Emergency visits (↓18%)	Privacy
Education	AI-learning	Dubai Smart 2030	Satisfaction (95%)	Citizen participation
Security	Edge, quantum	SAIoT-SL	Enhanced security	Complexity

presents key smart living solutions across domains such as AI-powered healthcare, AI-integrated education, and secure smart home infrastructure, highlighting their technologies, real-world implementations, outcomes, and persistent challenges.

3.6. Smart People

The smart people dimension of smart cities emphasizes citizen empowerment, education, skill development, and personalized services through the integration of AI, IoT, and digital twin technologies. Understanding citizen perceptions of AI is fundamental to designing acceptable, human-centered smart city services. A recent study identified five key citizen concerns regarding AI adoption: privacy fears, the desire for effortless benefits, discomfort with AI mimicking humans, avoidance behavior if perceived risks are high, and overall trust in AI technologies [61]. Incorporating these insights into the design of AI-enabled urban services is essential to foster acceptance and participation. These platforms often use sentiment analysis models, such as transformer-based language models fine-tuned on multilingual social media and feedback data, to adapt civic interfaces and detect public opinion trends in real time. AI-powered IoT systems further support intelligent city applications across domains such as healthcare, transportation, and civic engagement [62]. Recent innovations highlight the role of AI in enhancing cardiovascular health monitoring through atrial fibrillation detection combined with skin hydration sensing, the deployment of edge intelligence for low-latency operations in next-generation railway systems, and sentiment analysis for evaluating civic service performance based on multilingual social media data [62]. These applications reflect the growing ability of AI systems to not only automate functions but also to understand and react to human sentiments and behaviors. In parallel, the advent of 6G technologies has introduced the concept of integrating Network Digital Twins (NDT) into AI-based architectures, enabling real-time closed-loop optimization of complex communication networks [63]. These digital twins bridge the physical and digital worlds, enhancing urban mobility, communications, and public services. Cyber-resilient digital twin frameworks have also been developed for Internet of Vehicles (IoV) environments, showing significant improvements in security, latency, and energy efficiency [64].

Furthermore, AI-driven tools for online education, skill tracking, and career guidance using recommendation systems, clustering, and predictive modeling algorithms distinct from static digital dashboards are becoming critical components of smart living environments. Personalized career recommendation systems powered by machine learning models enable citizens to upskill in areas like data literacy, cybersecurity, and AI development, aligning workforce capabilities with the evolving demands of smart economies. A notable example is Singapore’s SkillsFuture initiative, which, as of 2024, has expanded its AI-powered platforms to provide personalized learning pathways, adaptive upskilling recommendations, and predictive labor market analytics. The underlying recommendation engine is typically built on collaborative filtering and deep learning-based profiling, using datasets comprising prior skill assessments, course engagement logs, and future labor market projections. The SkillsFuture AI platform uses machine learning models to suggest optimal training courses based on individual skill gaps, career interests, and future job market trends, thereby enabling citizens to maintain lifelong learning trajectories aligned with the evolving digital economy.

Table 6. People-centric AI initiatives.

Initiative	Technologies	Example	Impact	Concerns
Engagement platforms	AI chatbots	SkillsFuture	Job alignment	Data privacy
Digital health	AI IoT	Cardiovascular monitoring	Early diagnosis	Transparency
Network digital twins	AI networks	6G, IoV	Reduced latency	High costs

summarizes major people-centric AI initiatives, highlighting technologies such as chatbots, digital health systems, and network digital twins, alongside their real-world applications, societal impacts, and ongoing concerns.

To offer a synthesized overview of how AI is operationalized across the six smart city domains—governance, economy, mobility, environment, living, and people—Figure 4 visualizes the thematic focus areas alongside representative global case examples. Building on this, a synthesized overview of AI integration across the six smart city domains discussed above,

Table 7. Comparative summary of AI applications across smart city pillars.

Pillar	Key AI Applications	Technologies Used	Representative Cities	Benefits	Challenges
Governance	E-governance, citizen services, real-time decision-making	AI chatbots, blockchain, predictive models	Estonia, Singapore	Transparency, efficiency, enhanced service delivery	Data privacy, algorithmic bias, explainability gaps
Economy	Smart buildings, startup ecosystems, Industry 4.0	IoT, reinforcement learning, smart contracts	Barcelona, Shanghai	Innovation, energy savings, economic growth	Digital divide, policy hurdles, implementation cost
Mobility	Traffic prediction, MaaS, autonomous vehicles	CNN, GRU, Edge AI, blockchain	London, Helsinki	Reduced congestion, multimodal integration	Infrastructure cost, safety concerns, trust issues
Environment	Energy forecasting, air/water monitoring, climate alerts	IoT sensors, ensemble ML, edge computing	Amsterdam, Delhi	Emission control, disaster resilience, sustainability	Data quality, technical complexity, maintenance
Living	Smart homes, predictive healthcare, AI learning platforms	Federated learning, AIoT, digital twins	Toronto, Dubai	Personalization, improved well-being, energy efficiency	Privacy risks, system compatibility, scalability
People	AI skill training, citizen engagement, civic platforms	NLP, sentiment analysis, network twins	Singapore, Toronto	Inclusion, digital empowerment, adaptive governance	Literacy gaps, trust barriers, technology access

presents a comparative summary. It highlights key AI applications, enabling technologies, representative global case cities, expected benefits, and common challenges within each thematic pillar.

4. Global Case Studies and Comparative Insights

To highlight the global diversity of AI-enabled urban transformation, this section presents nine symbolic case studies of smart cities. Each was selected to represent a unique combination of geographic region, technological focus, and policy approach. The chosen cities, spanning Asia, Europe, North America, and the Global South, collectively capture various urban experiences across the six smart city pillars. This intentional diversity provides a balanced lens for comparing AI deployment strategies, governance models, and societal impacts. Figure 5 offers a high-level comparison of city performance across key indicators such as adoption, infrastructure, engagement, and measurable outcomes, setting the stage for deeper city-specific insights.

Among the nine cities examined, Singapore is analyzed in greater depth to showcase governance-centric AI integration. As a global pioneer in smart nation development, Singapore offers a strong case for examining the institutional, technological, and ethical dimensions of AI deployment in public services. This deeper focus allows for a more nuanced understanding of how long-term digital policy, unified infrastructure, and citizen-centric design contribute to successful smart city transformation.

4.1. Estonia

Through its flagship e-Residency program located in Tallinn, Estonia has become a world leader in smart governance by digitilizing nearly 99% of public services and introducing secure digital identities. Estonia has significantly lowered bureaucratic inefficiencies, cutting administrative load by 70% and increasing public satisfaction by 82% by digitizing government services and providing safe digital identities to people worldwide. Artificial intelligence-powered chatbots and blockchain have made smooth, open interactions 24/7 between users and the government possible. The X-Road interoperability platform enables seamless data exchange between over 900 organizations, while AI chatbots handle thousands of queries daily with minimal human intervention. Moreover, Estonia’s dedication to digital trust and policy consistency has helped it to set the standard in open government creativity [32,65,66].

4.2. Singapore

Singapore is leading the way in AI-driven urban development because of its Smart Nation project. Deployed across over 70 government agencies, the “Ask Jamie” virtual assistant is one of its signature breakthroughs. By providing multilingual, adaptive services covering public health, transportation, and education, the system has raised citizen involvement to 93%. Through the Smart Dubai 2030 framework, Singapore also excels in integrating artificial intelligence into home education, therefore promoting lifelong learning and professional upskilling. Notably, its unified governance infrastructure has enabled seamless inter-agency collaboration. Challenges remain, particularly in maintaining transparency in algorithmic decision-making and preventing overcentralization of AI control. Its proactive use of AI in pandemic response such as contactless screening and adaptive hospital triage demonstrated the robustness of its cross-sectoral AI readiness. However, continued concerns exist regarding surveillance transparency and algorithmic explainability. The city-state shows how technology can fit with citizen needs and policy frameworks [33,36,67,68].

4.3. Barcelona

The 22@ Innovation District of Barcelona has become a key hub in the European AI and smart economy ecosystem. Hosting more than 1500 startups concentrating on digital urban services and green technology, the district adds about EUR 3 billion yearly to the local GDP. Strong public–private partnerships and city-wide urban transformation investments drive this economic engine. Barcelona has increased assistance in recent years to encourage projects for sustainable mobility and smart manufacturing based on artificial intelligence. AI-powered urban dashboards are also used to forecast energy needs and mobility flows in the 22@ district. Nonetheless, disparities remain in digital access between core and peripheral neighborhoods, requiring inclusive planning reforms [69,70].

4.4. United Kingdom, London

Transport for London (TfL), which utilizes artificial intelligence to predict traffic patterns, reduce congestion, and optimize traffic light cycles, illustrates London’s leadership in smart mobility. These initiatives resulted in an 8% reduction in CO₂ emissions and a 12% reduction in traffic delays. TfL increased its system in 2024 with real-time edge-based AI rerouting in suburban areas. Complementing these developments is London’s Open Data system, which improves openness and builds public confidence in digital government. London also uses AI-based pedestrian analytics to redesign urban walkways and optimize traffic calming zones. Yet, financial and regulatory constraints hinder faster rollout beyond central boroughs [71].

4.5. Helsinki

Helsinki’s urban transportation has been transformed by implementing Mobility-as-a-Service (MaaS) via the WIM app. This system combines public and private mobility choices—including buses, subways, and scooters—into one payment and scheduling interface. Among young people, the MaaS system has led to a 20% drop in car ownership. Especially, Helsinki included a climate footprint calculator in 2024 to push users toward more environmentally friendly paths. Helsinki’s MaaS governance emphasizes user control and data privacy, supported by cross-provider data agreements. However, adoption gaps among older populations and rural areas highlight ongoing equity challenges [72,73].

4.6. Amsterdam, The Netherlands

Using AI-driven energy management tools, Amsterdam emphasizes environmental sustainability. Through real-time energy load prediction and adjustment, these smart grids reduce peak electricity demand by 25%. By adding microgrid-level controls, Amsterdam in 2024 scaled this system to enhance local energy resilience. By integrating artificial intelligence and IoT into energy infrastructure, Amsterdam sets the standard for climate-adaptive urban technologies. Amsterdam also launched a participatory AI governance board in 2024, allowing citizens to review and suggest changes to AI systems used in environmental monitoring [35].

4.7. India, Delhi

By using machine learning to forecast pollution levels 72 h in advance, Delhi’s SAFAR (System of Air Quality and Weather Forecasting and Research) project enables citizens and hospitals to take preventive actions. Personalized air quality notifications through mobile apps helped to improve the system in 2024. Despite technological progress, Delhi faces systemic challenges that limit the effectiveness of AI interventions. SAFAR’s reach is concentrated mainly in central zones, with limited deployment in peripheral districts. Infrastructural differences and patchy sensor coverage lead to reduced forecast accuracy in certain areas. Additionally, while mobile notifications provide advisory alerts, lack of integration with transportation and health response systems constrains city-wide impact. These issues are compounded by inconsistent data governance and limited inter-agency coordination. But, Delhi struggles with ongoing issues like low digital literacy and scattered urban planning that restrict more general involvement of its citizens [74,75,76].

4.8. Toronto, Canada

With areas fitted with AI-driven diagnostics, telemedicine, and smart home integration, Toronto is leading the way in smart healthcare. Emergency hospital visits have decreased by 18% thanks to these developments. Based on behavioral and IoT sensor data, Toronto introduced predictive mental health modules in 2024. Though the technology was maturing, early implementation phases battled with public knowledge and involvement gaps. Since then, Toronto has piloted community-based onboarding initiatives to improve digital inclusion, with a focus on marginalized neighborhoods [34].

4.9. Dubai, United Arab Emirates

Most especially in education, Dubai’s Smart Dubai 2030 project integrates artificial intelligence into the fabric of urban life. From inside their smart homes, residents access tailored, multilingual e-learning tools, achieving a 95% satisfaction rate. Earlier stages of the project were criticized for giving technology top priority over citizen inclusion. Through increased community involvement and government changes, Dubai has made progress toward balancing creativity with user-centric design [77,78,79]. In 2024, Dubai expanded its smart governance framework to include AI ethics panels and cross-cultural content testing to ensure inclusivity across its diverse expatriate population.

4.10. Thematic Comparison Across Cities

A cross-case comparison of AI-enabled smart city initiatives reveals not only regional variation in priorities but also structural and institutional differences in how urban technology is visualized and governed. The selected cities span a broad geographical spectrum, covering North America (Toronto), Europe (Estonia, London, Amsterdam, Barcelona, Helsinki), Asia (Singapore, Delhi), and the Middle East (Dubai). This diversity offers a rich comparative lens, as each region brings unique institutional frameworks, cultural expectations, and digital maturity levels to their smart city approaches. By analyzing cities from both developed and emerging economies, the comparison highlights how geopolitical context shapes AI adoption strategies, infrastructure capabilities, and citizen engagement dynamics.

To better understand the strategic differences among these cities, we compare their AI governance approaches, infrastructure models, implementation priorities, and citizen engagement strategies in the following subsections. This structured analysis helps to uncover not only local strengths but also recurring tensions and trade-offs present in global smart city development. To consolidate the findings across the nine city examples,

Table 8. Comparative case studies of smart cities.

City	Initiative	AI Technologies	Metrics	Success Factors	Limitations
Estonia	e-Residency	Blockchain, AI	Efficiency (↑70%)	Digital ID	Global barriers
Singapore	Ask Jamie	AI chatbots	Engagement (↑93%)	Adaptive AI	Explainability
Barcelona	22@ District	IoT, AI	EUR 3B GDP/year	PPP	Digital inequality
London	Smart mobility	AI analytics	Emissions (↓8%)	Real-time data	Infrastructure costs
Helsinki	MaaS	IoT, blockchain	Car ownership (↓20%)	Integrated mobility	Public adoption
Amsterdam	Energy grids	AI, IoT	Peak reduction (25%)	Local management	Privacy
Delhi	SAFAR AQI	ML Forecasting	72 h accuracy	Public health action	Digital divide
Toronto	Smart health	AI diagnostics	Emergency visits (↓18%)	Predictive care	Acceptance
Dubai	Smart Dubai	AI education	Satisfaction (95%)	Citizen-centric	Initial exclusion

summarizes key AI initiatives, enabling technologies, performance metrics, success factors, and noted limitations, providing a comparative overview of global smart city implementations. To support this comparative discussion,

Table 9. Matrix of AI applications across smart cities by sector.

City	AI Domain	Model/Algorithm Type	Data Sources	Automation Level
Singapore	E-governance, service automation	AI chatbots (adaptive NLP), Multilingual AI	Citizen queries, government service logs	Semi-autonomous with real-time interaction, human-in-loop
Estonia	Digital identity, Public services	Blockchain-integrated AI, Rule-based verification	Identity documents, usage logs	High autonomy in verification, manual override possible
Barcelona	Smart economy, startups	AI-driven innovation clusters, recommenders	Startup ecosystems, economic data streams	Varies by service, AI used in diagnostics and forecasts
London	Smart mobility, traffic systems	Edge-AI, CNNs, GRUs, random forest	Traffic sensors, public transit telemetry	Fully autonomous rerouting, predictive control
Helsinki	Mobility-as-a-Service (MaaS)	Route optimizers, blockchain + IoT	App-based transport logs, payment systems	Medium automation, user choice emphasized
Amsterdam	Smart grids, energy optimization	Time-series forecasting, anomaly detection	Smart meter feeds, grid telemetry, weather data	Fully automated load prediction and dispatch
Delhi	Air quality forecasting (SAFAR)	Regression trees, ML ensembles	Environmental sensors, satellite data, traffic stats	Advisory-level automation, manual public response
Toronto	Smart healthcare, predictive health	Diagnostic ML, behavioral predictive models	IoT health sensors, patient records	Autonomous alerts + human clinical verification
Dubai	Smart learning, home education	Personalized learning algorithms, recommenders	Learning history, feedback scores	High automation, user-adaptive interfaces

presents a structured matrix mapping AI applications across case cities by sector, model type, data input, and level of automation. This helps to illustrate the diversity in technical deployment and operational sophistication.

4.10.1. Governance-Centric vs. Sector-Centric AI Deployment

Singapore and Estonia epitomize a governance-centric model, where artificial intelligence is deeply combined with administrative systems. Estonia’s e-Residency program offers digital identity and cross-border access to services, while Singapore’s “Ask Jamie” AI assistant supports real-time multilingual service delivery across more than 70 government agencies. These initiatives reflect mature e-governance ecosystems where AI is treated as a systemic enabler of public service transformation.

On the other hand, cities like Toronto and Barcelona use a “sector-centric approach” where AI is focused on certain areas of the city, like healthcare in Toronto and economic innovation in Barcelona’s 22@ district. These cities use AI to increase public value in specific areas instead of changing how they govern themselves. This difference suggests that cities with a long history of good digital governance use AI in a wide range of ways, while cities with strong private innovation ecosystems only use AI in a few key areas that have a significant impact.

4.10.2. Centralized Infrastructure vs. Participatory Models

There are also differences in how governments work and how citizens can get involved. The Smart Nation program in Singapore is very centralized, with AI being used in a top-down way by government agencies that work together. On the other hand, Amsterdam applies participatory governance, where people help to make decisions about AI policies that have to do with monitoring the environment and surveillance.

These differences exemplify opposing paradigms: one prioritizing operational efficiency and centralized control (e.g., Singapore), and the other accentuating transparency and democratic accountability (e.g., Amsterdam). Both models are context-sensitive, reflecting differing political cultures and levels of civic trust. While centralized AI governance can offer consistency and efficiency, it also raises ethical concerns related to surveillance, data monopolies, and limited public oversight. Cities like Amsterdam counterbalance this risk by establishing citizen-led AI ethics panels, promoting algorithmic transparency, and enforcing digital sovereignty principles to preserve public trust. In contrast, models lacking participatory checks risk deepening power asymmetries and undermining digital rights.

4.10.3. Risk Tolerance and Regulatory Readiness

Cities also differ in how much they are willing to take on technological risks and how ready they are for AI regulation. In the area of smart mobility, London and Helsinki are good examples of how to do things differently. London uses edge-AI tools to improve traffic light cycles and cut down on traffic jams. This is a more experimental and infrastructure-heavy use of the technology. Helsinki, on the other hand, uses Mobility-as-a-Service (MaaS) systems that focus on integrating different modes of transportation and improving user experience.

London’s model focuses on predictive analytics and being able to respond quickly, while Helsinki’s model puts accessibility and social equity at the top of its list of transportation system goals. This suggests a range of methods, from technically aggressive infrastructure optimization to mobility planning that puts the user first.

4.10.4. AI for Sustainability and Public Health

Lastly, the reasons for using AI are different in each city. Amsterdam focuses on sustainability with AI-optimized smart grids and microgrid controls, with the goal of making the environment more resilient in the long term. Delhi, on the other hand, uses machine learning models to predict air quality and send out public health alerts.

Both cities use AI to deal with environmental problems, but Amsterdam’s approach is proactive and focused on infrastructure, while Delhi’s is reactive and focused on health. These differences are due to differences in the institutions’ capacity, the resources they have, and the policies they follow. The diverse approaches and institutional contexts highlighted above contribute to varying outcomes across cities. Figure 6 summarizes the key success factors and common challenges observed across the nine global case studies, capturing recurring patterns in AI implementation, governance readiness, and citizen engagement.

Across global smart city implementations, several common success factors have emerged. Leading cities such as Singapore, Estonia, Barcelona, and London have demonstrated that a strong government vision combined with policy continuity was fundamental for smart city success [32,33,70,71]. Singapore’s Smart Nation initiative and Estonia’s e-Residency program exemplify how consistent, long-term strategies have helped to accelerate AI adoption and digital infrastructure maturity [65,66]. Cities with citizen-centric designs, such as Singapore’s multilingual AI-based virtual assistants (“Ask Jamie”) [67] and Estonia’s seamless digital services, achieved higher levels of citizen engagement and trust [33]. Furthermore, proactive public–private partnerships fueled innovation; Barcelona’s 22@District project, supported by initiatives like Barcelona Activa, fostered over 1500 AI and green tech startups, contributing significantly to local GDP [69]. Open data policies and transparent digital governance, demonstrated in London’s Open Data Platform [71], further enhanced citizen trust and stimulated ecosystem innovation. Strategic investments in smart infrastructure, like Amsterdam’s AI-based energy grids [35] and Toronto’s Smart Health Clinics integrating healthcare with smart homes [34], led to substantial operational efficiencies, sustainability gains, and improved citizen outcomes.

However, common challenges were also evident. Fragmented or short-term planning notably impacted cities like Delhi, where despite pioneering initiatives like SAFAR’s AI-based air quality forecasting system [74,76], the lack of integrated infrastructure limited systemic impact. Similarly, low citizen digital literacy and awareness, particularly in Delhi and Toronto’s early phases, reduced citizen engagement despite available technologies [75]. Technology-centric approaches without adequate citizen focus, observed in Dubai’s initial Smart Dubai 2021 phase [77,79], led to criticisms around inclusivity and public value realization. Furthermore, the absence of comprehensive ethical and regulatory frameworks in some cities raised concerns about data privacy, transparency, and algorithmic fairness, particularly where AI systems expanded faster than governance mechanisms evolved [36]. Persistent digital inequalities between urban centers and peripheral areas further limited equitable access to smart city benefits, especially in emerging economies. In conclusion, cities that demonstrated the highest levels of success combined visionary leadership, citizen-first service design, robust open data policies, early investments in digital infrastructure, and responsible AI governance. Conversely, failures predomainantly stemmed from fragmented planning, insufficient regulatory oversight, low citizen engagement, and digital divides. These comparative insights provide a foundational understanding of best practices for future AI-enabled smart cities.

4.11. Cross-City Reflections and Limitations

While the comparative analysis reveals commonalities—such as the central role of governance capacity and citizen engagement—it also highlights sharp divergences affected by local context (

Table 10. Common smart city successes and failures.

Success Factors	Cities	Common Failures	Affected Cities
Vision, policy	Singapore	Fragmented plans	Delhi
Citizen-centric design	Helsinki	Tech-first design	Dubai (early)
PPP	Barcelona	Digital literacy gaps	Delhi, Toronto
Transparency	Estonia	Ethical gaps	Delhi

). Cities like Singapore and Estonia benefit from cohesive digital policy regimes, whereas Delhi and Nairobi face systemic barriers that limit the effectiveness of AI interventions. These imbalances suggest that smart city frameworks cannot be universally transplanted; they must be adapted to socio-political and infrastructural realities. Moreover, there remains a risk of techno-solutionism, where cities adopt AI without addressing deeper governance gaps or social inequities. Future research must therefore move beyond descriptive success stories to critically assess who benefits, who is excluded, and under what institutional conditions AI in cities becomes truly transformative.

5. Challenges and Barriers

Building on the case studies and thematic findings, this section critically analyzes three overarching barriers to AI implementation in smart cities: data privacy, ethical governance, and infrastructure inequality. It draws from both global initiatives and localized failures to explain recurring roadblocks. Figure 7 provides a visual representation of how these challenges are layered and interdependent.

5.1. Data Privacy and Security

As smart cities increasingly integrate AI, IoT, and big data technologies, concerns regarding data privacy and security have become central. The vast data flows generated by sensors, devices, and user interactions raise issues around surveillance, consent, and misuse [80]. For instance, China’s implementation of facial recognition technologies in public spaces exemplifies an aggressive surveillance model [81], contrasting sharply with Europe’s GDPR framework that prioritizes citizen data rights and mandates explicit consent [82]. Centralized data architectures are more likely to be hacked, so we need to look into decentralized solutions like blockchain-based ones [83]. Federated learning and other new models have been suggested as ways to strike a balance between smart city functionality and strong privacy protections by processing data locally instead of collecting it all in one place [84]. Therefore, finding a way to balance data-driven innovation with people’s right to privacy is still a significant problem for governments. For example, Toronto’s Sidewalk Labs project faced cancellation in part due to public backlash over unclear data governance and surveillance concerns. In India, several smart city projects have been criticized for collecting facial and biometric data without adequate consent frameworks, raising questions about citizens’ digital rights in emerging economies.

5.2. Ethical and Social Implications

Using AI systems in city government has raised many important moral questions. AI algorithms can have biases that make them make unfair decisions in important areas like policing, healthcare access, and welfare distribution [85]. Also, automation caused by AI has led to the loss of traditional jobs, especially in the logistics, transportation, and administrative fields. If upskilling programs are not put in place right away, this change could make social inequalities worse [26]. In the United States, algorithmic policing tools have been shown to disproportionately target minority communities, reinforcing historical biases. Similarly, in London, concerns were raised when AI-driven welfare screening systems flagged vulnerable families without sufficient human oversight, prompting an inquiry into automated decision transparency. Also, the digital divide is a significant problem for inclusivity because people who do not have access to high-speed internet or AI-driven services may not be able to take advantage of smart city benefits. Addressing these concerns demands not only technical improvements like bias mitigation and explainable AI but also robust participatory governance models that prioritize fairness, transparency, and citizen empowerment. Cities like Amsterdam have responded by establishing AI ethics councils that involve civil society in algorithm review processes [35,79], promoting more inclusive governance frameworks.

5.3. Infrastructure and Investment

The success of AI-driven smart cities hinges on the availability of robust infrastructure, including high-speed broadband, distributed sensor networks, cloud–edge architectures, and secure data storage facilities [86]. However, a considerable gap persists between developed cities, such as Singapore and London, and many developing cities, where investment in digital infrastructure remains insufficient [87]. For instance, in Nairobi and Dhaka, patchy broadband coverage and aging grid systems have significantly delayed the deployment of smart utilities and transportation platforms. These cities highlight how basic infrastructure readiness is a prerequisite for any AI-layered transformation. Funding constraints exacerbate this divide, particularly when public sector budgets are stretched thin. Private sector involvement through public–private partnerships (PPPs) has been a key enabler for infrastructure development, though it introduces new governance challenges concerning transparency and equitable service distribution [88]. Additionally, many urban regions struggle to modernize legacy systems, slowing the adoption of AI technologies necessary for smart transformation [89]. Barcelona’s efforts to expand its AI-driven urban mobility systems into outer neighborhoods have been hindered by legacy transit infrastructure, revealing the limits of retrofitting older systems to support real-time analytics. Without strategic infrastructure planning and inclusive funding mechanisms, smart city visions risk becoming exclusive projects, benefiting only a select segment of the urban population.

5.4. Risks and Failures in AI-Driven Smart Cities

AI technologies can bring about significant changes, but using them in cities has also come with significant risks and problems. Several cities have had problems with projects because of mistakes in ethics, infrastructure, or governance (

Table 11. Challenges to smart city implementation.

Category	Issues	Cities	Solutions
Privacy	Surveillance, consent	China, Toronto, India	GDPR, blockchain, federated learning
Ethical implications	AI bias, automation	US, London, Amsterdam	Explainable AI, ethics panels
Infrastructure	Digital divide, legacy systems	Nairobi, Dhaka, Barcelona	PPPs, infrastructure planning

). For example, Toronto’s Sidewalk Labs project was stopped in 2020 after many people complained about privacy issues and unclear rules for surveillance. Also, San Diego’s smart streetlight program, which used AI-powered cameras, was put on hold after people protested against unclear surveillance practices and a lack of public consent.

Algorithmic decision-making in public services has also had effects that were not planned. The courts in the Netherlands shut down an AI-based welfare fraud detection system called SyRI because it unfairly targeted low-income and immigrant neighborhoods. These examples show how important it is to have stronger laws, more open design processes, and more clear algorithms to make sure people are held accountable and trust the system.

There are still gaps in implementation, even in cities with advanced technology. In Dubai and Singapore, worries about bias and explainability in AI-based hiring or citizen scoring systems have led to discussions about the moral limits of using automation in government. These examples show that relying too much on technical efficiency without thinking about ethics or receiving input from the public can hurt the legitimacy of smart city projects.

Even though adding AI to smart cities has made governance, mobility, and resource use more efficient, there are still some major problems. Several city-level implementations have shown ethical and operational problems, often because they relied too much on technocentric models and did not have enough public oversight. For instance, Toronto’s Sidewalk Labs project was stopped in part because people were angry about unclear data governance and fears of surveillance. This shows what can happen when there is not enough citizen consent and transparency. People are worried that AI-powered police and welfare systems in London are making algorithmic bias worse and making structural inequalities worse. Also, many cities, especially those in the Global South, do not have the infrastructure, skilled workers, or cybersecurity skills to use AI on a large scale. This often results in fragmented use, data silos, and system weakness. Smart services could also leave some people out because they do not know how to use technology or do not have access to high-speed internet. These risks show how important it is to use AI responsibly, adapt governance models, and keep assessing risks to avoid outcomes that are unfair or not long-lasting.

6. Future Directions

The evolution of AI-driven smart cities is poised to accelerate dramatically through the adoption of emerging technologies, progressive governance frameworks, and interdisciplinary research initiatives. As illustrated in Figure 8, this transition will rely on the convergence of real-time AI capabilities (e.g., edge computing and 5G), ethical policy models, and cross-sector collaboration aimed at ensuring long-term sustainability and resilience.

6.1. Emerging Technologies

A primary emerging technology is the integration of Edge Artificial Intelligence (Edge AI), which facilitates real-time analytics and decision-making directly at the source of data generation. Unlike traditional cloud-centric architectures, edge AI minimizes latency, enhances data privacy, and improves energy efficiency—critical factors for autonomous vehicles, smart surveillance, and predictive urban maintenance [80,84]. Cities like Seoul are already piloting edge AI at scale in traffic management systems to predict congestion and reroute vehicles dynamically within milliseconds.

Furthermore, the proliferation of 5G networks is set to be a game-changer, offering ultra-high-speed and low-latency connectivity, thereby unlocking advanced AI applications in sectors such as healthcare, transportation, and energy [87]. Beyond these individual technologies, the synergy between AI, IoT, and blockchain architectures presents a promising avenue for building secure, decentralized, and intelligent smart environments, particularly for services like decentralized energy markets, verifiable citizen identities, and autonomous supply chains [82,83]. This convergence will be instrumental in achieving fully autonomous urban services with reduced risk of single points of failure.

6.2. Policy and Government Frameworks

However, technological advancements must be paralleled by robust policy and governance frameworks to ensure responsible AI deployment. The OECD AI Principles and the upcoming EU AI Act emphasize the importance of ethical design, algorithmic transparency, accountability, and human oversight [82,83]. Future smart city strategies must prioritize value-based AI governance, integrating fairness, inclusivity, and explainability at the core of urban AI systems [26].

One notable trend is the increasing emphasis on citizen participation in AI governance, where residents are involved in decision-making processes regarding surveillance, automated policymaking, and urban data monetization [85]. For instance, the city of Amsterdam has launched citizen-centered AI panels where locals can vote on major AI policies affecting urban surveillance and environmental monitoring. Participatory models such as Citizen AI Review Boards or Urban AI Consultation Panels are gaining traction globally, offering a pathway toward more democratic and transparent smart city developments. Establishing international cooperation on ethical standards and transparent auditing systems will also be necessary to prevent “AI ethics washing” by large urban projects.

6.3. Research Opportunities

The research landscape also presents several critical opportunities. There is a pressing need for longitudinal studies evaluating the long-term impact of AI on urban sustainability goals [81,85]. While AI-driven optimizations promise reductions in energy usage and traffic congestion, the environmental costs of training large AI models, deploying extensive sensor networks, and maintaining always-on smart systems remain underexplored.

Moreover, AI for disaster management and climate resilience is an emerging research frontier. Predictive AI models combined with IoT sensor networks can enhance early warning systems, optimize evacuation routes, and support post-disaster recovery efforts [80]. Pilot projects in Japan and California using AI-based earthquake prediction illustrate the tremendous potential but also highlight challenges around model reliability and public communication.

Another key research dimension is the exploration of AI ethics in urban contexts, encompassing algorithmic bias in housing allocation, public resource distribution, and law enforcement predictive analytics [26,85]. The growing digital divide between technology-rich and technology-poor urban areas also necessitates greater focus on equitable technology deployment, ensuring marginalized communities are not left behind in the AI-driven urban future. To complement these projections,

Table 12. Future directions for AI-driven smart cities.

Emerging Technology	Application Areas	Pilot Examples	Key Benefits
Edge AI	Traffic, analytics	Seoul, London	Real-time insights, reduced latency
5G Networks	IoT, autonomous vehicles	Singapore, Amsterdam	High connectivity, fast data transmission
Blockchain–AI synergy	Secure identity, marketplaces	Estonia, Amsterdam	Enhanced security, transparency
Ethical governance	AI regulation, citizen panels	Amsterdam, EU AI Act	Citizen trust, inclusivity
Quantum computing	Secure communications	Experimental (China, US)	Advanced encryption, computational power

outlines key emerging technologies, their application domains, pilot cities, and associated benefits that are shaping the future landscape of AI-driven smart cities.

7. Discussion

Though artificial intelligence is being adopted quickly in smart city sectors, significant discrepancies still exist in both research and practice. Most artificial intelligence applications stay siloed, concentrating on narrow functions like traffic forecasting or chatbot service delivery, even while present research and implementations show encouraging use cases in governance, mobility, energy, and healthcare. City systems are not well integrated to allow for whole, adaptive urban intelligence. Future studies have to investigate dynamic decision-making in complicated urban ecosystems as well as cross-domain interoperability of artificial intelligence. One important theme missing is the lack of longitudinal studies. Though few look at the long-term social, environmental, or institutional consequences of artificial intelligence in urban settings, most studies show short-term gains, e.g., energy savings and congestion reduction. Research should give multi-year, cross-city comparative studies of AI effects on sustainability, inclusion, and governance quality top priority. The creation of strong ethical frameworks suited to smart cities is another research area.

Although worldwide standards like the OECD AI Principles and the EU AI Act [90,91] offer high-level direction, many localized governance systems are not prepared to manage AI-specific hazards including real-time surveillance, algorithmic opacity, or biased service distribution. Urban-specific regulatory sandboxes, algorithm audit systems, and participatory design processes involving underprivileged communities in artificial intelligence policy creation are all necessary. There are still technical holes as well. Most AI implementations still rely on cloud-centralized systems, which increases questions about latency, privacy, and resilience. However, their scalability and dependability in crowded metropolitan areas need more empirical proof; edge AI and federated learning present interesting options. The environmental expenses of artificial intelligence training and deployment, particularly in relation to sustainability objectives, are also insufficiently investigated. The digital divide, at last, is still a constant struggle. Many developing countries lack the infrastructure, knowledge, or policy support to achieve comparable advantages even while cities like Singapore and Helsinki show effective large-scale AI use. Future work has to emphasize flexible, resource-aware artificial intelligence systems able to run in low-connectivity or resource-constrained settings without exacerbating current inequalities. All things considered, the sector has to advance from pilot projects and limited deployments toward integrated, inclusive, ethically based artificial intelligence systems operating across time, space, and social settings. Smart city researchers and practitioners have to co-develop multidisciplinary agendas combining technical progress with civic responsibility to guarantee the long-term legitimacy and social utility of AI-driven urban innovation.

7.1. AI Applications in Tackling Urban Flooding and Heat Island Challenges

The increasing frequency of extreme weather events due to climate change has boosted the urgency of developing resilient urban infrastructure. The most pressing climate-induced challenges facing cities today are urban flooding and the urban heat island (UHI) effect. Recent advances in Artificial Intelligence (AI) have facilitated data-driven, real-time solutions that address these phenomena’s physical and socio-environmental dimensions. In this context of urban flooding, AI models have been successfully used to model and predict flood damage at fine spatial resolutions. For instance, Ref. [92] developed a multivariable machine learning framework to assess flood impacts on residential infrastructure in coastal regions, showing the effectiveness of AI in localized risk assessment and real-world decision support. Similarly, Ref. [93] presented hydrodynamic models that estimate the vulnerability of wastewater treatment plants under extreme flood conditions, showing the importance of infrastructure strengthening as a key factor in urban sustainability.

AI also facilitates combining socio-demographic and spatial vulnerability data to mitigate flood risk. Ref. [94] demonstrated that a high significance level shapes community-level flood susceptibility, highlighting how AI-enhanced spatial models can uncover environmental inequities often overlooked in traditional planning approaches. Like flood management, AI is revolutionizing how cities comprehend and react to urban heat island effects. We used high-resolution datasets and predictive AI algorithms to map how temperatures change in dense urban areas. Ref. [95] provides detailed information about Manhattan’s thermal landscape, showing heat differences on a small scale. In addition, Ref. [54] used real-time weather forecasts and surface temperature models to support cities’ plans in a way that considers climate. Furthermore, Ref. [55] presented a conceptual framework that combines environmental risk assessment with social susceptibility analytics, laying the foundation for equitable urban heat mitigation strategies. These studies demonstrate AI’s transformative power in modeling, predicting, and alleviating climate-related urban hazards. As cities increasingly face the double threat of flooding and heat stress, AI delivers scalable, adaptive, and context-sensitive tools to support urban resilience while promoting climate justice. Future smart city initiatives should prioritize including these AI-driven models into infrastructure planning, early warning systems, and policy frameworks to ensure sustainable, inclusive urban environments.

7.2. Quantitative Impact Summary

The smart city case studies we looked at showed that using AI in different areas of the city has led to significant quantitative results. Estonia’s e-Residency program has cut down on bureaucratic work by 70% and made citizens happier by more than 80% in terms of governance [32]. The Ask Jamie chatbot in Singapore says that more than 93% of users are engaged with it across more than 70 government agencies.

Barcelona’s 22@ innovation district adds about EUR 3 billion to the city’s GDP every year in the economy pillar. AI-enabled smart buildings have seen a 99.1% increase in efficiency, and China’s smart grids have cut forecasting errors to less than 1% [45]. Helsinki’s MaaS systems have cut car ownership by 20% and made it easier for people to switch between modes of transport. London’s AI traffic systems have cut traffic by 12%.

Amsterdam’s smart grids cut peak load by 25%, Delhi’s SAFAR system gives 72 h accurate air quality predictions, and Toronto’s predictive AI clinics cut emergency health visits by 18%. These are all good for the environment and public health.

These numbers, which are also shown in Table 8 and Table 9, show that AI has a measurable, multidimensional effect on how well smart cities work. But, differences in how local governments work, how developed infrastructure is, and how involved citizens are show that strategies need to be flexible and aware of their surroundings.

8. Conclusions

The use of AI in smart city frameworks has led to a huge change in many areas of city life, including governance, the economy, transportation, the environment, housing, and giving citizens more power. The case studies looked at include Estonia’s pioneering e-Residency program, Singapore’s Smart Nation initiative, Barcelona’s startup ecosystem, and London’s AI-enabled traffic management. They show that AI can improve operational efficiency, sustainability, inclusiveness, and quality of life when used wisely. Comparative analysis shows that cities with participatory governance mechanisms, strong digital infrastructure, and well-defined ethical oversight frameworks (like Singapore and Helsinki) have done better than those that only focus on technology and do not have enough people involved. However, making cities that are truly smart and fair is not easy. This challenge highlights the need to view urban AI systems not as neutral technologies but as socio-technical constructs shaped by institutional context, political will, and cultural values. A city’s AI maturity is as much about ethical readiness and civic capacity as it is about data pipelines and model performance. Data privacy and security concerns are still very important, especially with the rise of more powerful surveillance tools. AI bias, digital inequality, and job loss are all ethical problems that need careful management and design that includes everyone. The digital divide becomes worse when there are gaps in infrastructure maturity between developed and developing areas. This disparity also cautions against uncritical replication of AI models across cities. Smart city strategies that work in highly digitized contexts may fail in lower-resource environments unless adapted to local realities. Instead of exporting frameworks wholesale, there is a growing need for adaptive, context-sensitive AI governance that aligns with regional needs and capacities. Successful models show that three things are very important for success: clear data governance, strong citizen participation, and public–private innovation ecosystems. On the other hand, technocentric strategies that do not take into account social acceptance and ethical safeguards often lead to failure. Emerging technologies such as edge AI, 5G, and AI–blockchain synergies are already shaping experimental use cases in cities like Amsterdam and Dubai, particularly in energy optimization and secure data management. These innovations promise to enhance real-time analytics, decentralization, and data protection, though their outcomes remain at an early stage. At the same time, the emergence of global AI governance frameworks such as the OECD AI Principles and the EU AI Act underscores the growing institutional push for algorithmic transparency, accountability, and citizen empowerment in AI deployment. It will also be important to perform interdisciplinary research that looks at AI’s long-term effects on society, disaster resilience, and sustainable urbanization. In the end, the success of smart cities will depend not only on technological progress but also on the growth of policies that put people first, ethical AI ecosystems, inclusive infrastructures, and governance models that let people take part. Ultimately, the future of AI in urban development must move beyond efficiency metrics toward questions of justice, trust, and collective agency. This requires interdisciplinary collaboration between technologists, urban planners, legal scholars, and citizens themselves to co-create futures that are not only smart but also sustainable and socially grounded. Cities of the future can only be smarter, more sustainable, more resilient, and more fair if they take a balanced, value-driven approach.