Geospatial artificial intelligence (GeoAI) powered by large language models (LLMs) now enables natural‑language querying, generation, and interpretation of spatial information, supporting autonomous GIS workflows. This capability introduces governance challenges that standard AI ethics discussions miss, such as passive location inference from mobility traces, bias amplification driven by spatial autocorrelation and scale effects, hallucinated spatial facts, and compounded uncertainty across multimodal inputs.
We identify eight recurring issues in LLM‑enabled GeoAI: data provenance and consent, spatial privacy and inference risk, algorithmic bias and spatial inequity, spatial mechanisms as structural risk—including spatial autocorrelation, the modifiable areal unit problem, and scale effects—LLM‑specific technical risks, lack of explainability, policy and regulatory gaps, and public enablement and workforce development. For each issue we describe the underlying mechanism, cite a representative literature example, and assess the current response ranging from largely unaddressed to actively debated or partially regulated.
Building on this synthesis, we propose a governance‑aware architecture for autonomous GIS that maps each issue to enforceable controls and auditable artifacts throughout the geospatial data lifecycle. A flood‑response routing scenario illustrates how privacy audits, bias mitigation, and explainability modules can be embedded at data collection, model inference, and result dissemination stages.
The review highlights a persistent evidence gap: most proposed controls remain conceptual, with few field‑tested evaluations of their effectiveness.
Consequently, future research should prioritize empirical validation, development of spatially specific interpretability tools, and training programs aligned with these emerging risks.
Review