Most City Digital Twins Aren't Twins Yet
An investigation of 99 urban digital-twin initiatives exposes the gap between sophisticated city models and operational reality, revealing why municipal leadership and reliable data may matter more than artificial intelligence.
By By The Spatial Signal Editors
· Week of November 9, 2026 · 8 min read
Listen · shortened audio edition

Image created by AI
A city can now recreate entire neighborhoods in extraordinary three-dimensional detail, overlay streets with traffic information, and use artificial intelligence to simulate what might happen next.
But does that mean it has a working digital twin?
In August 2026, Rancho Cordova, California, announced that Niantic Spatial had reconstructed approximately four square kilometers of the city using drone imagery. The resulting model was detailed enough to read street signs and measure physical distances.
Project partner Human Machine Collaboration Institute (HMCI) was already reporting traffic simulations using the reconstructed environment. Yet the broader task of integrating those capabilities into routine municipal planning remained a work in progress.
A digital twin becomes more than a digital replica when it changes not just what a city can see, but what a city can reliably do.
The distinction matters. Creating an accurate digital representation of a city is one achievement. Using it to improve decisions about real streets, utilities and infrastructure is another.
As technology companies increasingly describe digital twins as living, intelligent or even agentic, a fundamental question confronts local governments: how much of that promised intelligence has actually reached municipal operations?
An investigation published earlier this year offers a revealing answer.
The 99-project reality check
In March 2026, researchers from RWTH Aachen University and TU Dortmund University published an examination of 99 urban digital-twin initiatives across Europe, North America and Asia.
The results revealed a substantial gap between technological ambition and implementation.
Of the 99 projects, 56—approximately 57%—were classified as research, development or prototype initiatives rather than systems in routine operation.
The researchers also assessed technical maturity. Thirty-nine initiatives were categorized as real-time-data models, while 15 had reached two-way integration, allowing information to flow between digital and physical systems. None reached the study's highest category: full automation.
The findings require context.
Approximately three-quarters of the initiatives examined were European, with a particularly strong concentration in Germany. Only 12 were based in the United States. The researchers also conducted municipal interviews in summer 2024, before several American projects now attracting attention.
The study was exploratory rather than a representative worldwide census. Its results cannot establish that 57% of all urban digital twins are nonoperational today.
Nor does the absence of automation mean that a project has failed. A model that helps planners evaluate a proposed development can deliver considerable value without independently controlling infrastructure.
Nevertheless, the findings challenge the assumption that increasingly sophisticated digital models are automatically becoming intelligent municipal management systems.
The missing ingredient may not be technology
One of the investigation's most consequential findings concerned something much less fashionable than artificial intelligence.
Municipal involvement.
Across the projects examined, more active participation from local authorities tended to accompany more advanced implementation. In approximately 35% of cases, authorities were actively involved in developing the twin or incorporating it into their work.
The relationship was observational, not proof that government participation directly caused success. Better-resourced cities might simply have been more capable of achieving both.
But it points toward an obstacle that impressive demonstrations can obscure.
A digital twin may combine GIS databases, LiDAR surveys, engineering models, traffic cameras, utility records and sensor networks. Those sources are often managed by different departments, maintained to different standards and updated at different intervals.
The researchers' municipal interviews identified familiar problems: fragmented data, poor accessibility, coordination difficulties, shortages of staff and funding, and concerns about information security.
A sophisticated visualization platform cannot resolve those institutional weaknesses on its own.
Someone must maintain the data. Someone must determine which records are authoritative. Someone must decide how the system's outputs affect engineering or operational decisions.
And someone must remain accountable when the digital representation differs from conditions on the ground.
These questions become more important, not less, as digital twins acquire greater analytical capabilities.
America's cities are moving forward—but at different speeds
Recent American projects demonstrate both how much the technology has advanced and how difficult it remains to prove sustained operational benefits.
In Raleigh, North Carolina, the city worked with Microsoft, NVIDIA and Esri to bring video feeds from 12 traffic cameras into an AI-assisted digital environment.
According to an August 2026 Microsoft customer case study, the project moved from concept to pilot in six weeks.
The goal was to help transportation officials better understand intersection activity and evaluate changes to traffic-signal timing.
Microsoft reported that Raleigh expects potential commuter savings of $9.7 million from improvements in traffic flow.
That figure is a projection, not an independently verified financial benefit already achieved. It illustrates the potential economic value of even modest traffic improvements while leaving the actual outcome to be demonstrated.
Colorado Springs offers a different perspective.
Using federal SMART grant funding, the city tested an adaptive traffic-management system incorporating radar, LiDAR, video analytics and a digital model of intersection operations.
Field testing took place at two intersections between April 2024 and January 2025. Subsequent validation simulations indicated potential reductions in traffic delay ranging from 15.8% to 23.7% compared with conventional coordinated signal control.
The U.S. Department of Transportation's evaluation documents an encouraging technical result, but not the realization of those savings across the proposed 48-intersection expansion.
That distinction is critical. A successful simulation can justify further investment, but operational performance must ultimately be measured on the street.
Rancho Cordova illustrates another stage of the journey: creating a high-fidelity spatial foundation quickly and using it to test scenarios. The technology partners reported early traffic simulations, but publicly documented evidence of resulting improvements in municipal services remains limited.
These projects are not interchangeable. One is a rapidly developed pilot, another has completed federally evaluated field testing, and another is evolving from detailed reconstruction toward planning applications.
Together, they show that digital-twin progress is better understood as a spectrum of capabilities than as a simple divide between deployment and failure.
The most useful twin may not look like a twin
Some of the more convincing examples of digital modeling in government are also less visually dramatic.
New York City's Department of Environmental Protection has developed an internal digital model of its sewer system to investigate stormwater flooding.
In October 2025 testimony, the agency explained how problems in one part of the drainage network can contribute to flooding elsewhere.
Using a modeled storm scenario involving 2.15 inches of rainfall per hour, the department identified 86 locations with the greatest need for intervention.
The model helped establish where investment should be prioritized.
There is no public evidence in that testimony that the system is an autonomous, continuously updated digital twin. Nor does identifying priority locations establish that flooding has subsequently been reduced.
But the example reveals an important distinction between technological sophistication and public value.
A targeted model of an underground network that supports capital planning may be more immediately useful than a photorealistic representation of an entire city without a defined operational purpose.
That does not diminish advanced digital twins. It clarifies what they ultimately need to accomplish.
The evidence is more complicated than a failure narrative
A separate investigation published in June 2026 examined 33 city-information-modeling and urban digital-twin initiatives worldwide.
Its findings complicate the assumption that these systems are primarily elaborate 3D maps.
Nearly 70% of the projects reported applications involving simulation, prediction and verification. Approximately 61% used 3D visualization for property and land administration, while 48.5% addressed urban planning.
Yet only 9.1% reported climate-adaptation applications.
These categories overlap, and reported capabilities do not necessarily establish regular operational use or measurable improvements.
Still, the comparison is important.
Digital-twin development is neither uniformly immature nor advancing evenly. Cities are using different combinations of spatial data, modeling and simulation to address different problems.
Some applications are already well established. Others remain experimental. And some of the most ambitious promises—particularly autonomous management of complex urban systems—have yet to demonstrate their reliability at meaningful municipal scale.
The useful question is no longer whether a city possesses something called a digital twin.
It is what that system actually does.
What cities should demand before investing
For local-government executives, the distinction has immediate financial and operational consequences.
A digital twin is rarely just a software purchase. Maintaining one can require recurring expenditure on data collection, integration, cloud computing, sensor maintenance, cybersecurity, personnel and licensing.
A demonstration may establish technical feasibility without revealing the cost of keeping the representation accurate over five or ten years.
That changes how municipal investments should be evaluated.
Can a proposed system use existing GIS and asset-management records? How frequently will its information be updated? Which department will maintain it? Can its underlying data be exported or used independently of the vendor's platform?
And, most importantly, what decision will improve because the system exists?
For a transportation agency, the answer could be fewer minutes of intersection delay. For a water utility, it might be reduced pumping costs or improved maintenance scheduling. For a planning department, it could be faster, more consistent development review.
These are measurable outcomes. A convincing 3D environment, by itself, is not.
As artificial intelligence becomes more deeply integrated into geospatial platforms, another distinction will become increasingly important: the difference between systems that recommend actions and those authorized to execute them.
A digital twin that predicts congestion is different from one that adjusts signals automatically. A model that identifies possible infrastructure defects is different from one that changes maintenance priorities without human review.
The more autonomous these systems become, the more demanding the requirements for reliability, security, transparency and accountability.
The next phase won't be measured in polygons
For years, digital-twin progress has often been demonstrated through the detail and scale of virtual environments: more buildings, more accurate surfaces, richer textures and increasingly realistic simulations.
Those advances matter. Accurate geometry and dependable spatial data are essential foundations for understanding the physical world.
But the next test is less visual.
It is whether cities can maintain the information, integrate it into established workflows and demonstrate that public services improve as a result.
The 99-project investigation provides an important warning, but not a verdict of failure. Urban digital twins are advancing, and useful operational applications already exist. What remains uneven is their transformation from ambitious projects into sustained municipal capabilities.
Artificial intelligence may accelerate that transformation. It may also expose weaknesses in underlying data and governance that cities have not yet resolved.
The most revealing measure of success will not be how faithfully a digital twin reproduces a city's roads, buildings and underground infrastructure.
It will be whether the people responsible for those assets can use it to make better decisions—and prove that those decisions made a difference.
A digital twin becomes more than a digital replica when it changes not just what a city can see, but what a city can reliably do.