The Default

Two startups, two sales decks, identical structure. Same first slide. Same second slide. Same third. The logos were different. Everything else was the same. Both had been built with the same AI design tool, and neither client had any particular reason to notice what they had accepted — because the output looked right. It looked like something a designer made. It looked finished.

That is the moment worth paying attention to. Not the sameness, which is a design problem. The moment worth paying attention to is what happened before the sameness: someone handed a task to a tool, received an output, and moved on. The output was convincing enough that the question of how it got there did not arise.

That is not a design story. That is a trust story.

What a Default Actually Is

Every AI system has defaults. The palette it reaches for when you do not specify one. The structure it builds when the prompt is open-ended. The choices it makes in the absence of instruction. These are not neutral — they are trained preferences, baked in from the data and the design philosophy of whoever built the system. When a tool's outputs share a recognizable aesthetic across thousands of different users and tasks, you are seeing those preferences made visible at scale.

The interesting thing about a default is that it is most powerful when it goes unnoticed. A default you question is a decision point. A default you accept is just the output.

Most people accept the output. Not because they are incurious, and not because the tool has deceived them. They accept it because the output is plausible, the task is done, and there are other things to do. The cognitive work of interrogating an AI output competes with everything else pressing for attention — and a polished, confident-looking result tends to win that competition by a wide margin.

The Effort the Output Obscures

The designers who produce distinctive work with these same tools are not using different technology. They are doing something different before and during the generation: mood boards, hand sketches, deliberate friction placed between themselves and the first plausible result. They are, in effect, refusing to let the default function as a decision. They are staying in the process long enough to make choices that are actually theirs.

That requires effort. It requires knowing what you want before you ask, which means having developed enough taste to recognize when what you got is not it. And it requires a tolerance for the discomfort of not being done yet — of staying with uncertainty, as one designer quoted in a recent New Yorker piece put it, long enough to find something new.

Most people do not want to do that. Not because they are lazy, but because the output looks good. Good enough is a threshold, and AI tools have raised that threshold considerably. When the floor of what you can produce without effort is higher, the motivation to go further drops. That is the mechanism. It is not about trust in the system being too high. It is about the gap between the effort required to evaluate the output and the effort required to accept it being so large that acceptance becomes the rational choice.

Where the Gap Opens

This is the same dynamic that shows up in higher-stakes contexts — in medical decisions, in security assessments, in financial analysis. The form is different. The structure is the same. An AI system produces a plausible output. The output looks authoritative. The user, under some combination of time pressure, cognitive load, and confidence in the tool, accepts it without fully engaging with what it contains or how it was produced. Trust moves faster than comprehension. The gap opens.

In those contexts, we talk about the gap as a safety problem. In design, we talk about it as an aesthetic problem — a homogenization of the visual web, a flattening of creative distinctiveness. But the underlying dynamic is the same thing. The output displaced the process. The user accepted what the tool decided.

The question of whether that matters depends on what is at stake. For a sales deck, the cost is visual sameness and a missed opportunity to communicate something distinctive. For decisions with higher stakes, the cost of not staying in the process is harder to recover from.

What Noticing Requires

There is a design principle embedded in the tools that produce good outcomes here: the best systems do not just deliver outputs. They create moments where the user has to engage — where trust cannot be extended without some active act of interpretation. Small friction, deliberately placed. Not enough to frustrate, but enough to interrupt the automatic acceptance response.

That principle applies beyond interfaces. The users who produce something genuinely their own with AI tools are the ones who have built that friction into their own process. They do not evaluate the output when it arrives. They have already done enough thinking, before the prompt, to know what they are evaluating it against.

The default is only invisible until you name it. Once you see it, you have a choice. But the seeing requires having stayed in the process long enough to recognize when the output does not actually reflect what you wanted — which is harder to do than it sounds when what you got already looks finished.

That is the real cost of defaults: not the outputs they produce, but the questions they foreclose.