Artificial Intelligence in Urban Planning: A Bibliometric Analysis and Hotspot Prediction

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript offers a thorough and timely bibliometric analysis of using artificial intelligence (AI) in urban planning. The research is methodologically robust. It offers development trends in planning practice and the future of using AI in urban planning.

The paper contributes to contemporary literature. It demonstrates how AI has transformed the practice of urban planning. Combining bibliometric analysis with AI-assisted literature review and predictive modeling is new.

 The manuscript is structured properly. The flow is logical. The graphs are clear and detailed. The research questions are clear. However, I would add a third question. A (How) question might provide an executionable component to the agenda for future research. Adding a third question helps in framing the problem space. The references are relevant to the topic and are up to date.

Nonetheless, the manuscript can be improved by first indicating how generative AI outputs were checked or filtered. Text in some figures, such as Fig. 1, need to be clear and bold to be informative. Some keywords are already in the title, such as urban planning. It is recommended to replace them with other keywords to give the paper better exposure on the web search engines. The limitations of the paper can benefit from a more critical-self assessment. Finally, a roadmap for future research can put the authors’ recommended actions into motion.

A good paper that I enjoyed reading. However, there is room for improvement.

Comments on the Quality of English Language

Some sentences are too long, vague or awkward, requiring revision for clarity.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors, it is my pleasure to review your article.

Have you really manually reviewed 4,594 records? (line 92)
In figure 2 you mention high frequency >5, and then high frequency 1-2, I am assuming it was supposed to be low frequency.

The main question of the research is what are the current trends, research hotspots, and future directions for the application of AI in urban planning? The authors aim to answer this by combining bibliometric analysis, generative AI-assisted literature synthesis, and machine learning-based keyword prediction using a Support Vector Machine  model. As mentioned in my earlier review, i belive that the topic is highly relevant and timely. AI's role in urban planning has been expanding rapidly, yet I have not come across many studies that provide a comprehensive, large-scale bibliometric analysis, especially combined with predictive modeling to forecast research hotspots, thus this paper addresses methodological as well as knowledge gap.   Unlike prior sector-specific reviews or traditional bibliometric studies, this paper introduces a three-stage developmental framework of AI in urban planning, while appliying generative AI tools to synthesize insights from nearly 4,000 publications.   As mentioned earlier the claim that 4,594 documents were manually reviewed (line 92) appears unlikely. Clarification is needed regarding the extent of manual vs. AI-assisted review. Additionally, there is an inconsistency in terminology, namely “high frequency” is defined both as >5 and as 1–2 across different sections. This needs correction (e.g., Figure 2 and Table 1).   Meanwhile, conclusions are well-supported by the data. The transition from early to mature phases of AI adoption is substantiated by longitudinal keyword analysis, publication volume, and technological mapping. The SVM-based prediction of future research directions logically extends from the bibliometric data and supports the stated research questions. References are relevant and up-to-date.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Review land-3806584-peer-review-v1

Date: 20/August/2025

Title: Artificial Intelligence in Urban Planning: A Bibliometric Analysis and Hotspot Prediction

 

1. Original Submission
   • Recommendation

Major Revisions

 

2. Comments to Author:

Ms. Ref. No. land-3806584

 

 

2.1 Overall and general recommendation

Summary.

The manuscript presents a bibliometric and thematic mapping of how artificial intelligence is used in publications related to urban planning. It reports temporal trends, leading countries and institutions, frequent keywords, and application areas (with a strong emphasis on transport/mobility). The abstract also promises a “technical application guide for urban planning.”

 

Assessment.

The topic is timely and potentially useful. However, the paper does not sufficiently connect AI to the core tasks of urban planning—i.e., the spatial configuration of the city (form and land use), multi-scale accessibility, provision of services, and the socio-economic and environmental processes that drive urban growth and change. Much of the analysis remains descriptive (counts, co-occurrence networks) without explaining what problems in urban planning AI is helping to solve, with what data, and with what limits.

 

Recommendation.

Reconsider after major revisions.

 

The paper has potential if it

• defines an operational framework for “urban planning,”
• separates data sources/infrastructure from AI techniques,
• actually delivers the promised application guide (problem → data → method → outputs → limits), and
• strengthens methodological transparency for the bibliometric mapping.

 

This can become a useful, practitioner-oriented map of how AI is actually used in urban planning if you ground it in planning practice (spatial + socio-economic), clarify the taxonomy (data vs. methods), deliver the promised application guide, and make the bibliometric workflow reproducible.

 

 

 

2.2 Major Comments

• Please provide a concise framework of what “urban planning” covers in practice: spatial structure and morphology of the city, land-use and zoning, multi-scale accessibility and network structure, service/location standards, socio-economic drivers of growth and decline, housing and land markets, environmental quality (air/noise), and climate-related risks. Without this, bibliometric counts do not translate into substantive insight.

 

• Urban planning is inherently spatial. Geographic Information Systems (GIS), spatial statistics, and approaches such as space syntax are the backbone for analyzing form, accessibility, and behavior. Show more clearly how AI complements these tools (e.g., land-use classification from imagery, accessibility prediction, urban morphology metrics, spatial interaction models), and reflect that consistently in figures (where GIS currently appears marginal or is absent).

 

• Separate data/infrastructure from AI methods: throughout the paper, data sources and infrastructures (IoT, remote sensing, “big data”) are presented as if they were AI techniques. Please distinguish clearly:
  • Data layers: remote sensing, sensor networks, administrative records, surveys.
  • Platforms/infrastructure: IoT networks, clouds, pipelines.
  • AI techniques: machine learning, neural networks, graph learning, reinforcement learning, time-series models, (optionally) explainable AI (XAI), etc.
    This will make Figures 12–14 and related tables conceptually consistent.

 

• Deliver the promised “technical application guide.” The abstract commits to a guide, but none is provided. Add a compact matrix mapping urban-planning problems to data, AI method, evaluation metrics, typical outputs, and limits, for example:
  • Land-use/land-cover mapping;
  • Growth/expansion simulation and scenario testing;
  • Accessibility and service catchments;
  • Housing supply/demand and price dynamics;
  • Environmental risks and resilient design (flood/earthquake/wildfire risk, multi-hazard exposure, evacuation and shelter siting);
  • Green/blue infrastructure planning and urban heat island mitigation;
  • Low-carbon and energy-aware planning (district energy, solar potential);
  • Equity and inclusion (exposure to pollution, access to amenities).
    Each row should cite representative studies.

 

• Coverage is skewed to transport: transport/mobility is important, but it is only one part of urban planning. Expand the taxonomy to include socio-economic (housing, land markets, inequality), environmental (UHI, air quality, green structure), and risk/resilience strands, and reflect that in the keyword clustering and discussion.

 

• Smart city + AI work routinely raises data governance questions (data stewardship, purpose transparency, public oversight, participation). Rather than a generic ethics paragraph, we suggest a short governance subsection linked to planning practice: who owns/curates the data, how models are validated for policy use, and how planning agencies and communities are involved in interpreting outputs.

 

• Raw counts by country/institution are hard to interpret. Please (i) normalize (e.g., by population, GDP, R&D spending), (ii) map the actual places of study (cities/regions analyzed), and (iii) discuss differences by city type (size, density, region, income level). Avoid causal inferences about national policy from unnormalized publication counts. That is, country-level counts need normalization and more useful geography.

 

 

• Urban planning benefits from teams that span architecture/urbanism and data/engineering. Consider reporting simple interdisciplinarity indicators (co-authorship across fields, department types, funding sources) to show whether the field is actually bridging that gap. We believe that measuring interdisciplinarity in some way can provide a lot of information about the application of AI to solve urban problems.

 

• Tables and figures must be reproducible and informative.
  • Tables 7 and 8: state precisely how they were constructed (selection criteria, coding rules, data sources). As is, they miss the chance to guide practitioners.
  • Figures 12–14: align with the clarified taxonomy (data vs. method), remove technologies not developed in the text, and justify links.
  • Use vector formats, increase label sizes, and add self-contained legends.

 

• Document search strings, databases (e.g., Web of Science/Scopus), time window, inclusion/exclusion criteria, de-duplication, disambiguation, and at least one sensitivity check (e.g., alternative queries, database comparison).

 

• Close each research question with an explicit answer, then include a limitations section (topic coverage, language/database biases, taxonomic ambiguities, etc.) and specific future work.

 

• To avoid conflating overall field growth with AI adoption, report the share of urban planning papers that use AI over time (AI subset vs. total urban-planning output). That is, we recommend measuring the relative growth of AI within urban-planning literature.

 

                      

2.3 Minor Comments

 

• If terms like XAI (explainable AI) are kept, define once and use consistently. Avoid introducing new labels that are not used in the corpus.
• Ensure GIS appears where appropriate. In ecosystem diagrams, show GIS as a cross-cutting analytical layer for spatial tasks.
• IoT/remote sensing/big data: treat these as inputs or infrastructure, not AI per se; when they appear in figures, label their role explicitly (data acquisition/ingestion).
• Graphic quality: export figures in vector format (PDF/SVG), improve contrast and font sizes, and provide clear legends.
• Align the promised “technical guide” in the abstract with a concrete section in the results/discussion; avoid repeating lists without added interpretation.
• When you associate a technique with a planning problem, cite one or two representative studies and note required data and typical metrics.
• Shorten long sentences, remove vague phrases, and keep the discussion focused on actionable insight for planners.

 

 

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf