When everything looks the same: AI, the flat design déjà vu, and why real-world thinking wins

We have been here before

A little over ten years ago, flat design swept through the industry. It started as a genuine correction, a move away from glossy buttons and fake leather towards clarity, and for a while it felt fresh. Then it became a template. Within a couple of years, apps, websites and even brand identities looked interchangeable: the same long shadows, the same pastel palette, the same icons in the same circles. Clients began asking for things that looked like everything else, because everything else was what they saw.

The trend did not fail because flat design was wrong. It faded because sameness always does. When everything looks alike, nothing stands out, and the work that people remember is the work that broke the pattern for a good reason.

Why AI pushes everything towards the middle

Generative tools learn from what already exists and return what is most likely. That is their strength and their limit. Ask for a landing page and you get the average of thousands of landing pages: competent, polished and instantly forgettable. The more people use the same tools with similar prompts, the more the results converge, and the faster a fresh idea becomes a cliché.

None of this makes the tools bad. It simply means that the output starts where the crowd already is. Anything that sets the work apart has to come from somewhere else.

Where the difference comes from

In my experience, it comes from the real world. From watching how people actually use a service, at a counter, on a bus, in a hurry. From understanding the market a product competes in, the habits of its users and the constraints nobody writes down. From a client’s history, their brand, their quirks. And, for me, from the hours spent drawing, painting and making things by hand, which train the eye in ways a screen never will.

Real-world thinking produces solutions that look the way they do because of the problem, not because of a trend. A public portal designed around the people who queue for it will not look like a start-up template, and it shouldn’t. A brand built on a client’s actual story will not resemble the generated average, because the average has never heard that story.

Using the tools without disappearing into them

I use AI every day, so this is not a call to abandon it. It is a reminder of where it belongs. I let it speed up the parts that should be quick, such as drafts, variations and prototypes, and I keep the decisions about character, tone and direction for myself. A generated layout is a starting point to react to, never a conclusion.

The test I apply is simple: if the result could belong to any other product, it is not finished. Something specific to this client, these users and this problem has to be visible, otherwise the work is just one more drop in the same sea.

The pattern will break again

Flat design gave way to richer, more expressive work once people grew tired of seeing the same thing everywhere. The same will happen with the generated look, probably sooner than we expect. When it does, the designers who kept their feet in the real world, who looked, listened and made things by hand, will be the ones whose work still feels alive. That is not nostalgia. It is simply how taste has always worked.