AI made competent design free, and in doing so it destroyed competent design as a competitive advantage. Everyone has a nice logo now. What is scarce, and therefore valuable, is evidence that a human made a decision.
Something genuinely changed in the last two years, and most of the commentary about it has been either defensive or triumphalist. Both are missing what actually happened.
AI did not make design worse. It made competent design free.
You can now generate a logo that is well-proportioned, technically clean, and appropriate to its category in about ninety seconds, for effectively nothing. Ten years ago that outcome cost a few thousand dollars and took two weeks. That is a real capability and pretending otherwise is silly.
But here is the consequence nobody in my industry wanted to sit with: competence was most of what a lot of design services were actually selling. And when the floor rises for everyone, standing on the floor stops being a position.
Everyone has a nice logo now. A nice logo is therefore worth nothing.
The thing that is actually scarce
Walk a category today and you can feel the effect. A lot of brands look more polished than they did five years ago and less distinguishable from each other. The average went up. The variance collapsed.
Which means the scarce good is no longer execution quality. It is evidence of a decision.
A machine can render an idea beautifully. What it cannot do is stand in front of a brewery owner, understand that the thing making them different is the fact they refuse to distribute outside the valley, and decide that the whole identity should be built around a boundary rather than a product. That is not rendering. That is judgment about what matters, made by someone who understood the business.
And judgment leaves fingerprints. That is what people are responding to when they say something feels human - not the wobble in the line, but the sense that somebody chose this specifically, for this business, and could tell you why.
Three signals people are actually reading
The market has developed a fairly reliable, mostly unconscious set of tells. Worth naming them, because they are what your prospects are running on whether they can articulate it or not.
Specificity. Does this mark contain something that could only belong to this business? A reference to the actual place, the actual process, the actual person. AI output tends toward category-appropriate rather than business-specific, because it is generalising from the category. Specificity is the clearest possible signal that someone did the work of understanding.
Deliberate imperfection. Visible linework, hand-drawn elements, organic forms, a texture that could not be a default. Not sloppiness - chosen irregularity. In a feed of machine-smooth output, the thing with a discernible hand in it carries a trust premium, and that premium is now measurable in what people will pay.
Coherent restraint. The tell of generated work is often additive: a bit of gradient, a bit of texture, a bit of everything that reads as current. Restraint requires knowing what to leave out, which requires knowing what the thing is for. It is the hardest quality to fake because it is defined by absence.
None of these are about craft skill in the traditional sense. They are about the visible residue of thinking.
Where this gets misread
Two failure modes, both common right now.
Performing handmade. Adding a wobbly line to a fundamentally generic mark. Rough-textured typography on a brand with nothing rough about it. This is the same generic output wearing a costume, and people detect it quickly - it reads as a trend rather than a truth. The imperfection has to be connected to something real about the business or it is just another convention.
Refusing the tools entirely. The point is not that AI is off-limits. It is very good at exploration, variation, and burning through the obvious ideas fast so you can get past them. Using it to sketch is fine. Shipping its first coherent output as the answer is the problem, because its first coherent output is by definition the most average solution in the space.
The distinction is whether a human made the decisions or ratified them.
What this changes commercially
If you buy brand identity, the calculus moved.
The cheap option is genuinely better than it used to be, and for some purposes it is now sufficient. A holding company that will never compete on shelf, an internal tool, a temporary campaign - generated work is a reasonable answer and I would not argue otherwise.
But if your brand has to be chosen over something else - on a shelf, in a pitch, on a lawn sign, in a category where a buyer is comparing four options - then looking professional is table stakes and being specific is the entire game. That is the situation where the gap between generated and considered work is not aesthetic. It is commercial.
I made this argument about Realtors and I will make it again more broadly: in a crowded category, sameness reads as risk. AI has made sameness cheaper and more accessible than it has ever been. Which means the cost of taking the cheap route is now higher, not lower, than it was - because everyone else took it too.
The reframe
For most of the last decade, hiring a designer was substantially about buying execution you could not produce yourself. That justification is weakening and it is not coming back.
What you are buying now is a decision you cannot make yourself - because you are too close to the business to see which part of it is actually the interesting part, and because deciding requires giving things up, which is much easier for someone who is not emotionally attached to any of it.
The rendering is the cheap part now. It always kind of was. We just used to be able to charge for it.