When most people hear about artificial intelligence reshaping cities, their minds leap to sci-fi visuals: robot traffic controllers, AI-designed skyscrapers, or virtual reality planning simulations. Those headlines are eye-catching, but beneath the surface of flashy renderings and automated building permits, something more subtle—and more transformative—is happening.
Across the world, AI is beginning to influence the DNA of cities not through spectacle, but through data, decision-making, and democracy. In other words: how cities know themselves, how they make choices, and how people can have a say in shaping the future. And while this story isn’t making the front page yet, it’s the one that matters most.
Cities That Learn Without Stealing Data
Imagine your neighborhood wants to track air pollution or traffic stress—but you don’t want a big tech company scooping up your personal data in the process. That’s where a clever approach called federated learning comes in. Instead of collecting everyone’s data in one central system, federated learning trains AI on local devices—like your phone or a community-owned sensor on a lamppost. The raw data stays with you, but the system still learns. It’s like a neighborhood school where every student keeps their own notebook, but the class still advances together. This idea is already being tested in places where trust in institutions is fragile or where surveillance concerns run deep. In practice, it could mean that citizens collectively monitor noise, heat, or air quality—and AI helps them make sense of the patterns, without ever invading privacy.
Urban Memory, Supercharged
Most cities have long and tangled histories: masterplans, protests, zoning rules, forgotten promises. But these documents usually live in dusty folders or hard-to-navigate databases. What if AI could pull those scattered bits together and speak urban history fluently? Enter a new kind of artificial intelligence that acts like a knowledge translator. Imagine an urban planning chatbot trained not just on laws and maps, but also on community feedback, climate studies, and decades of local plans. When asked about a new development proposal, it could tell you, “Actually, there was a similar plan proposed in 1997—and residents raised concerns about flooding then too.” These tools, which use something called retrieval-augmented generation, are still in their infancy. But they could be powerful allies in public meetings, helping citizens ask smarter questions and understand the stakes. More importantly, they make sure the city doesn’t forget.
Who Owns Urban Intelligence?
As AI gets smarter, the big question isn’t just what it can do—but who it serves. Cities need to move from being consumers of AI (buying off-the-shelf tools) to governing it like public infrastructure. One promising approach is the use of model cards—documents that explain how an AI was trained, what its limits are, and how it might go wrong. Think of it like a nutrition label for algorithms. Even more exciting? Some cities and researchers are experimenting with community-written model cards. This means residents get a say in defining what counts as fairness or harm, before AI decisions are made. Alongside that, urban data trusts are emerging—legal entities that don’t just store data but steward it, acting in the public’s interest. Imagine a cooperative that manages data about vacant buildings and decides, with community input, how it’s used. Instead of tech companies calling the shots, it’s a civic board with a mandate for equity and transparency.
AI That Listens Better, Not Louder
Public participation in planning is often frustrating. Meetings are boring, the loudest voices dominate, and feedback gets lost in bureaucratic fog. But what if AI could help listen better—not just faster? New tools can sift through thousands of comments, emails, or voice recordings and identify key themes, emerging concerns, or overlooked perspectives. They don’t just measure sentiment (“happy” or “angry”)—they trace arguments, values, and evidence. This approach, called argument mining, is like giving planners a superpower to make sense of complex, messy public input. It also flips the script: instead of inviting residents to react to pre-baked plans, planners can co-create the agenda based on what people are already saying. Think of it as turning the comment box into a live feed for city-making.
Not Sci-Fi—Just Smarter Cities
All of this might sound futuristic. But these tools exist now—they’re just not in every planner’s toolkit yet. That’s partly because they’re not as flashy as 3D renders or AI-generated facades. But they matter more. They’re about how cities know themselves, how they change, and how they can be more just, more participatory, and more alive to the people who call them home. AI won’t replace urban planners or architects. But it might just make them better listeners, better learners, and better partners in the dance of city life. The question isn’t whether AI will shape our cities—it’s whether we’ll use it to amplify democracy, or drown it out with automation.
The good news? The quiet revolution is already underway. And it’s inviting all of us in.
by Markus Appenzeller
pictures created with the help of ChatGPT
