Feb. 9, 2026 update: I updated this article to add my response to something posted on X that was, frankly, ill-informed about taste and AI. So I added my response to the top, with my original article below. Combined, they express my thoughts more fully.

Creativity and taste is not a moat. It’s a differentiator.



All AI models have taste. But since most are trained on accumulated vs. curated design, the resulting taste is a reversion to the mean.



Infinite minds make taste worse. You need a refined, subjective selection of training data. Quantity ≠ quantity.

---

Original article:

For a designer, taste is both 100% subjective and also 100% observable as a skill. Taste is just the curated set of decisions a designer used to make something.

Taste is hard to measure objectively in terms of good and bad, since taste is highly subjective from one designer to the next. (Of course, you can always measure taste relative to yours.)

However, taste IS very measurable in terms of alignment vs. non-alignment, because you can count the number of designers that align with a specific taste or aesthetic. (That is how design movements such as the Bauhaus form, or in more recent times, Apple's industrial design.)

And because you can measure alignment, you can absolutely improve your taste as a skill by refining your curated set of design decisions.

A practical example: Font pairings. Obviously, someone with zero taste can just randomly pick two fonts and call it a day. But you can also refine your decision making for fonts that work well together, either by copying a designer you deem to have “good” taste, or creating a personal rubric based on what you know about typography, balance and visual design. Or both.

All this to say, big-box LLMs will always struggle to define taste because it predicts a probable outcome acceptable for a majority of designers, rather than a refined statement that aligns to your personal taste.

If you want AI to be a better reflection of your taste, you need to curate your decision set and add it as knowledge or a skill into your AI-assisted design process.