Brand guidelines are dead. Long live the .md.

Everywhere right now, brand guidelines are being generated from a prompt. A hundred pages, neatly structured, visually flawless, and finished in a few minutes.

That is impressive. But it also shows how interchangeable these documents have become.

Among teams using AI in their design system, 60 percent use it to create documentation. Right after code, it's the most common application.

That is understandable. Design system teams are chronically understaffed. 61 percent say they don't have enough people. So the first thing to get automated is what costs a lot of time and often has little effect in day-to-day work: documentation that gets created because it belongs to the system.

If a machine can write your central brand rulebook in a few minutes, that says little about the machine — and quite a lot about the rulebook. It describes a brand, but it doesn't keep one alive.

92 percent of design systems have documentation. Only 45 percent of teams are satisfied with it. So the problem isn't that documentation is missing. The problem is that it sits too far away from the actual work.

Outdated on arrival

A brand doesn't stand still. New products appear, requirements change, additional channels arrive. Language evolves. Markets shift.

Even carefully crafted brand guidelines therefore only ever describe one specific moment in time. Yet many companies treat them like a final answer.

The dark mode is missing. There's no component for the new use case. The defined tone no longer works in a changed market. So decisions start happening outside the system. First as an exception, later as the norm.

The guidelines stay clean. The brand has long moved on.

Maintenance belongs inside the work

Anyone who works with design systems knows the routine: a design needs something that wasn't foreseen. A new component, an additional state, or a different weighting within the existing rules.

Then the alignment begins. The solution gets designed, tested, documented and, at some point, fed back into the system. Days or weeks can pass between decision and documentation. Sometimes it never happens at all.

AI doesn't make this work obsolete. But it can move it closer to the moment the decision is actually made.

When a new solution works in the product, it should be able to become part of the system immediately. Not after the next status meeting. And not in a document maintained separately from the product.

The system has to live where the work happens.

The brand moves into the project folder

Brand guidelines as a static document have had their day. Not because brands no longer need shared rules. But because a PDF is no longer the place where those rules do their work.

The visual layer remains: components, typography, colors, behavior and motion. What's new is the layer behind it. An .md that explains why the system works the way it works. Which principles stay constant. Where variation is allowed. How the brand behaves — not just how it looks.

This file lives right inside the project. Next to the code, versioned and changeable at any time. Designers, developers and AI systems all work from the same context.

A markdown file doesn't replace a component library. It doesn't replace design judgment either. It takes over the job a static manual increasingly fails at: making decisions available while the work is happening.

This website works exactly that way. Its design system lives as an .md in the same project as the code. Every change takes its cue from it. When a convincing solution contradicts the existing rules, the solution isn't automatically discarded. Maybe the rule has to change.

Not documentation about the work. A part of the work.

The design system stays visible. Its intelligence becomes readable.

The gap is the job

78 percent of the designers and developers surveyed by Figma say AI makes their work significantly more efficient. At the same time, only 32 percent rely on its results.

This difference is not a temporary shortcoming of the technology. It shows where the real responsibility sits.

AI can generate variants, apply rules, check relationships and update documentation. It can make deviations visible. But it doesn't reliably decide which exception is right, when a system becomes too tight, or why a design feels wrong despite following every rule.

That still takes people with experience and judgment. People who define what belongs to the brand, which deviation is the right one, and when a rule has to change.

AI makes production faster. That doesn't mean fewer decisions — it means more.

So brand guidelines don't disappear without a successor. They change format.

A document you look at becomes a file you work with.