You live in Astoria. You know the smell of souvlaki, the sound of the N train, the way the sun hits the Ditmars stop. Now watch ChatGPT turn it into a generic ‘hipster coffee shop with exposed brick.’ That’s the moment you realize AI doesn’t see you. It sees a stereotype.
Last week, writer Jack Maguire did something simple: he asked ChatGPT to visualize every neighborhood in New York City. The results were a parade of clichés. SoHo became black turtlenecks and brick walls. Williamsburg turned into flannel shirts and artisanal pickles. Harlem got jazz clubs and brownstones. Every image was flat, polished, and utterly wrong.
Generative AI isn’t hallucinating. It’s remembering. And what it remembers is the internet’s most boring version of you.
This isn’t a failure of technology. It’s a feature. The model is trained on billions of images scraped from the web—stock photos, Instagram posts, travel blogs, and real estate listings. The most common visual signal for ‘Williamsburg’ is not the actual people or the bodega on the corner. It’s the curated, marketable version that sells. The algorithm picks the median, the average, the safest bet.
And that’s dangerous. Not because AI is bad at art, but because we’re outsourcing our sense of place to a machine that only knows the median. When we let AI define our neighborhoods, we lose the friction, the grit, the lived experience that makes a city a city.
The real danger of AI slop isn’t that it’s ugly. It’s that it’s convincing. And once you’ve seen the algorithm’s version of a place, you’ll struggle to see the real one.
I asked a friend who grew up in Bed-Stuy. She laughed at the AI’s image of brownstones and stoops, then got quiet. ‘It’s like someone described my block to a robot who never visited.’ That’s the problem. The robot never visited. It only knows the postcards.
This is the uncomfortable truth behind every viral AI image generator: the output is a mirror. It reflects back the most stereotyped, most reductive version of our own culture. The internet has already flattened New York into a collection of Instagrammable tropes. ChatGPT just made them visible.
So next time you see an AI-generated New York, remember: it’s not a window into the city. It’s a mirror. And what it’s showing you is the dullest, safest, most marketable version of yourself. Don’t let it be the last word.
FAQ
Q: Isn't this just a tech glitch?
A: No, it's a deliberate feature. The model is optimized to produce the most common pattern in its training data, not the most accurate representation. The 'glitch' is in our expectation that AI should capture nuance when it's designed to average everything.
Q: What's the practical takeaway?
A: Don't use AI for cultural or geographic understanding. Use it for tasks where averages are fine—like summarizing a menu or generating generic stock imagery. For anything that requires lived experience, specificity, or nuance, trust humans, not algorithms.
Q: Could AI's blandness be a good thing?
A: Maybe. It could be a cure for our obsession with hyper-specificity. If every AI image of a neighborhood looks the same, it might remind us that most places are more alike than we admit. But that's a cold comfort when the richness of a city is exactly what makes it worth living in.