Speakers
Description
Introduction and Research Problem
Public participation is expected to democratise planning by bringing experiential, local, and marginalised knowledge into decisions that shape places. Arnstein (1969) argued that participation can remain tokenistic when it does not redistribute power. Its democratic quality therefore depends not only on who is invited, but also on who frames the problem, decides what knowledge counts, interprets public input, and links it to formal planning decisions. Digital tools can broaden participation, yet unequal access, privacy concerns, limited institutional capacity, and uncertainty about how public input is used remain unresolved. Artificial intelligence (AI) adds a new layer to this relationship.
Within participatory processes, AI can help participants express or translate ideas, visualise alternatives, process public input, moderate discussion, and support prioritisation. Based on 50 cases in 22 countries, the OECD (2026) describes tools that may improve accessibility and support participation at greater scale. Planning studies also identify both opportunities and risks (Du et al., 2024). Raymond et al. (2025) report existing and anticipated uses, including the potential analysis of citizen feedback, as well as perceived risks related to data protection, misinformation, verification, and handing decisions over to automated tools. In an adjacent field, Arsenault and Kreps (2026) found that large language model assistance made public-comment writing easier and improved externally assessed comment quality, but did not improve participants’ self-reported understanding of policy. Together, these studies suggest that AI may strengthen expression or information processing without changing who controls interpretation or formal decisions.
Research has examined the technological opportunities and ethical risks of AI in planning (Du et al., 2024; Sanchez et al., 2025). To our knowledge, however, existing reviews of AI in participatory urban planning have not systematically examined four issues together: what AI does, who controls its use, how participatory data are governed, and whether its use changes participants’ power in planning decisions. Relevant evidence is also dispersed across planning, civic technology, public administration, human–computer interaction, and AI governance.
This study therefore aims to critically map how AI is incorporated into participatory urban planning and what the available evidence shows about its implications for power relations. It addresses two questions: (1) At what stages and for what purposes is AI used, who selects, configures, operates, and oversees it, and whose participatory data does it process? (2) What evidence shows whether and how these arrangements affect three dimensions of power: participants’ voice; their control over how contributions are interpreted and connected to decisions; and the distribution of benefits and burdens?
The review will trace the use of AI across the participation process, from the formation of public contributions through algorithmic processing to a formal institutional response or planning decision. It will distinguish claimed democratic benefits from effects supported by observed changes in participation or decision-making.
Conceptual Framework
AI is treated not as an autonomous decision-maker but as part of a socio-technical arrangement that includes models, interfaces, datasets, prompts, organisational routines, professional judgement, and institutional authority. Power is examined through three dimensions: voice, control, and the distribution of benefits and burdens. Voice concerns who can participate and whether their contributions are preserved. Control concerns who can select, configure, question, correct, or refuse a tool, as well as who decides how its outputs are interpreted and used in formal decisions. The distributive dimension asks who gains from greater accessibility, time savings, or analytical capacity, and who bears errors, misrepresentation, exclusion, privacy risks, or other burdens.
These dimensions connect planning debates on meaningful participation with participatory AI research. Using Arnstein’s ladder, Corbett et al. (2023) found that most studies they examined informed or consulted the public rather than sharing decision-making power. This review also examines how benefits and burdens are distributed.
Participatory data are not treated as neutral inputs: decisions about what is collected, made machine-readable, retained, reused, and converted into outputs shape whose knowledge counts. Taylor’s (2017) data justice framework identifies three pillars: (in)visibility, (dis)engagement with technology, and antidiscrimination. They guide the analysis of whose experiences become visible to AI, whether participants can question, correct, or refuse its use, and how benefits and risks are distributed among social groups. Data governance will be examined through data sources and consent, access and retention, reuse—including model training—traceability, and options for correction or deletion. The review also asks whether transformed contributions can be traced to formal institutional responses and planning decisions.
Methodology
This study adopts a critical scoping review design. It follows JBI methodological guidance (Peters et al., 2020) and will be reported in accordance with PRISMA-ScR (Tricco et al., 2018). A scoping review is suitable because this emerging field includes diverse study designs and application reports that are not suited to statistical meta-analysis.
Eligibility criteria follow the Population–Concept–Context framework. The population comprises citizens, community organisations, and organised stakeholder groups contributing to participatory planning. Planners, facilitators, public authorities, researchers, and technology providers are treated as institutional and technical actors whose roles in selecting, configuring, operating, and overseeing AI will be examined.
The concept is limited to AI systems used within a defined participatory urban planning process to help participants express, translate, or visualise ideas; classify, cluster, summarise, or prioritise their input; or support moderation and deliberation. Eligible systems may include generative and conversational AI as well as other AI-based tools for language or image processing. Eligibility depends on the system’s role in participation, not on a specific model type.
Eligible sources must document an implemented application or a prototype evaluated with citizens or organised stakeholders. Conceptual proposals, synthetic-data-only demonstrations, and systems that do not elicit or process participatory input will be excluded. Tools used only to provide information, predict or optimise spatial outcomes, conduct technical spatial analysis, or produce non-participatory visualisations are therefore outside the scope. An automated function will count as AI only if the source identifies an AI or machine-learning component. Reviews and general policy frameworks will support background and citation tracking, not outcome analysis. Grey literature will be eligible only when it documents a specific application.
The context is limited to urban and neighbourhood-scale spatial planning. Environmental or transport planning applications will be eligible only when they concern a defined spatial decision within a city. Regional planning and general public-policy consultations without a direct urban spatial component will be excluded.
Scopus, Web of Science, ACM Digital Library, PAIS Index, and TR Dizin will be searched from database inception through 31 August 2026 using English and Turkish terms. A bounded grey-literature search will cover selected repositories of the OECD, European Parliament, European Commission, and UN-Habitat. Google Scholar and backward and forward citation tracking will supplement these searches. The same eligibility criteria will apply throughout, and search strategies and screening decisions will be documented.
The review will be limited to sources available in English or Turkish because of the research team’s language capacity; the resulting limitations will be acknowledged. Two reviewers will independently screen titles, abstracts, and full texts and resolve disagreements through discussion. Selection will be reported in a PRISMA-ScR flow diagram. Documents describing the same application will be linked so that each application is counted only once.
A structured data-charting form will record the planning context, participant groups and access conditions, participation stage, AI type and function, and who selects, configures, operates, and oversees the system. It will also record whether participants use or help shape the system, verify its outputs, or only provide input; these roles may overlap. Where reported, the review will document data sources, consent, access, retention, reuse—including model training—and whether AI outputs can be traced to original contributions. It will also record human review, disclosure of AI use, options to opt out or request data correction or deletion, the ability to challenge AI interpretations, and reported effects on voice, control, benefits, and burdens across groups. Missing information will be coded as “not reported” rather than as evidence that a practice or safeguard was absent.
The analysis will combine descriptive mapping with critical thematic synthesis. Cross-tabulations will compare participation stages and AI functions according to who controls their use, how data are governed, and what effects are reported for voice, control, and the distribution of benefits and burdens. Claims will be grouped by the type of evidence supporting them: stated aims; outcomes reported by providers or practitioners; experiences reported by participants; changes observed in the participatory process; or documented changes in a formal decision or outcome. Stated aims and provider reports will not be treated as equivalent to observed or documented changes. Because the sources are diverse, a single quality score will not be applied to all of them.
Expected Results and Contribution
A preliminary scan points to an imbalance in the available evidence. Studies more often document what AI makes faster or easier—such as translation, visualisation, and the processing of large volumes of input—than whether it changes agenda-setting, accountability, or influence over final decisions. Many sources also appear incomplete about important governance questions unanswered: where participatory data come from, whether they are reused, who selects and controls the AI system, and whether participants can inspect, correct, or challenge AI-generated outputs. These observations remain provisional and form the basis for systematic examination.
At the centre of the analysis is whether benefits at one stage of participation carry through to later stages. An application may make it easier for people to express their views while leaving control over interpretation and formal decisions unchanged. Each application therefore will be traced, as far as the available documentation permits, from public contribution through AI processing to an institutional response or planning decision. Where sources do not establish a connection between these stages, this is recorded as a reporting or evidence gap rather than taken as proof that no connection exists. The analysis also considers whether applications that include traceability, human review, disclosure of AI use, and opportunities for correction provide clearer evidence of transparent and accountable treatment of public input. Following the full participation chain helps distinguish a larger volume of input from a genuine redistribution of power.
The intended contribution is a structured mapping and critical synthesis of where and how AI is used in participatory planning, who controls it, whose data it relies on, and what evidence supports its reported effects. The accompanying stage-sensitive framework traces the three dimensions of power—voice, control, and the distribution of benefits and burdens—across participation stages and AI functions. Comparisons also take account of actor roles, data-governance conditions, and the type of evidence supporting each claim. This structure enables comparison across different applications without treating technological innovation as evidence of democratic transformation. It also makes visible the geographical, institutional, and social gaps in the available literature. For planning practice, the framework offers a way to assess whose voice is strengthened, whose data are used, who benefits or bears risks, whether public contributions reach formal decisions, and who remains accountable for AI-mediated outputs.
References
Arnstein, S. R. (1969). A Ladder Of Citizen Participation. Journal of the American Institute of Planners, 35(4), 216–224. https://doi.org/10.1080/01944366908977225
Arsenault, A. C., & Kreps, S. (2026). Whose voice counts? The role of large language models in public commenting. Big Data & Society, 13(1). https://doi.org/10.1177/20539517261419341
Corbett, E., Denton, R., & Erete, S. (2023). Power and public participation in AI. In Proceedings of the 3rd ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization (pp. 1-13). Association for Computing Machinery. https://doi.org/10.1145/3617694.3623228
Du, J., Ye, X., Jankowski, P., Sanchez, et al. (2024). Artificial intelligence enabled participatory planning: a review. International Journal of Urban Sciences, 28(2), 183–210. https://doi.org/10.1080/12265934.2023.2262427
OECD. (2026). Artificial intelligence and the future of citizen participation: Typology of applications, opportunities and challenges for democratic innovation. OECD Publishing. https://doi.org/10.1787/a1ee2e0a-en
Peters, M. D. J., Marnie, C., Tricco, A. C., et al. (2020). Updated methodological guidance for the conduct of scoping reviews. JBI Evidence Synthesis, 18(10), 2119-2126. https://doi.org/10.11124/JBIES-20-00167
Raymond, C. M., Nummi, P., von Wirth, T., et al. (2025). Uses, opportunities and risks of artificial intelligence in participatory urban planning. Discover Cities, 2(1), Article 93. https://doi.org/10.1007/s44327-025-00137-4