How AI Is Reversing the Branding Process

First published in HORIZONT on September 9, 2026.

1. What is being produced for the brand with AI?

In most companies, presentations, proposals, social media posts and internal documents come out of AI tools like Copilot, Canva or the popular chatbots every day, including in departments that never used to design anything. It works the same way internally. Employees ask the internal assistant which template to use, how the logo should be applied, how to write about the company. The assistant answers with what it finds: the maintained brand portal if it's connected, otherwise the shared drive, and at worst a five-year-old presentation. Whether companies produce with AI is no longer the question. The question is what the tools take their cues from. If the answer is "from what the tool pieces together," the company produces a slightly different brand hundreds of times a day. Result: brand blur.

2. What AI makes of the brand when AI builds on AI

Systems increasingly hand work over to other systems: one agent gathers information, a second builds something from it, a third puts it out. At every handover, only what has been explicitly defined gets through. Anything that exists only as a feel in the heads of experienced colleagues gets lost along the way, and nobody notices. Each station interprets a little, small deviations add up, and in the end there's a result that nobody decided on in that form. Afterward, it's almost impossible to trace which handover changed it. Result: blur and arbitrariness once again, the very opposite of strategic brand management.

What this does to brand design and brand experience

For a person, variance is leeway. For a machine, variance is a problem. Take a brand whose guidelines describe its imagery as warm, natural, slightly desaturated. An image generator never hits that mood the same way twice. Tone of voice is no different. A tone described as humorous can mean anything from a subtle wink to a dad joke, and the text AI makes a new call every time. When one person selects the images or writes the copy, that hardly matters. With a thousand automatically generated applications, you get a thousand variants, all somehow within the frame, but none exactly as intended. That isn't the guideline's fault. Written for people, it has worked for decades. It just doesn't give the machine any orientation, because the description leaves too much open.

Design doesn't lose value as a result. It gets defined more strictly. Image mood as measurable values instead of a feeling. Tone of voice with examples and counterexamples instead of a single adjective. This isn't a return to the rigid manuals of the past: leeway can be defined too, as a range with clear limits. The dynamic brand systems of recent years show how it's done; their variety follows defined rules. That's why machines can handle them: they can deliver variance that follows rules, but not variance that's only felt. Brand guidelines that also work for AI tools are unambiguous and structured, written so a machine can read them.

This also changes what gets designed in the first place. Not everything can be solved with a better description; some elements the machine simply can't reproduce reliably. Just as print and screens once did, AI tools now feed back into design: anyone who knows this will design such an element differently or deliberately do without it. What emerges is a new discipline of definition, in which every element is chosen more consciously and specified more precisely than ever before.

The exception: just as important as the rule

The discipline of definition also has a flip side. You can only seriously decide what to hold back once you know everything that can be spelled out and automated. And that decision is what sets a brand apart: some elements stay in human hands even though the machine could easily produce them by now. The people behind the brand have decided: we're not handing this over.

A hand-drawn illustration, an in-person event, a personal reply instead of a generated text: the more the machine takes over in everyday work, the more value there is in the elements that clearly don't come from it. But that value doesn't create itself. The brand has to make this decision deliberately, stick to it and show it: we do this differently, on purpose. Without that signal, no one on the outside can tell whether something was deliberately left out or simply left over. And what looks left over looks outdated.

For strategy, this means it's no longer enough to describe what the brand stands for. What counts is how it differs from everyone producing with the same tools: what it defines precisely, and what it deliberately leaves out.

Upside down: brand development reloaded

The classic brand process puts the tools at the end: first you decide and design, then you codify the overall picture in guidelines, and only at the very end do people and systems apply the result.

That order is now reversing. An assessment becomes the first step: before anything gets defined, you find out how AI systems read the brand today and what the tools can reliably deliver. Anyone who starts a rebrand without this is designing for a production reality they don't know.

And the process needs an owner. When machines deliver, someone has to define what they're allowed to deliver and be accountable for the result. That isn't a technical role but a brand management role, and in most organizations nobody holds it today.

There's also an obligation from outside. Since August 2026, the EU AI Act has required machine-readable labeling of AI-generated content. You can treat that as a tiresome requirement, or as an opportunity to make your entire brand machine-ready.

What you should know about your own brand

Three questions are enough to get started:

How do the common AI systems represent our brand today? You can test this directly: ask the systems yourself, several times, in different ways. The result is often sobering and always revealing.

What are our own teams already producing with AI, and what are they taking their cues from? The honest answer shows how wide the gap between the documented brand and the lived brand already is.

Which of our rules exist only as interpretation in people's heads? Everything that lives there is invisible to machines and gets lost at every handover.

The answers tell you what to do next: which foundations need structure, which definitions are missing, and who decides from now on. Company size barely matters. What matters is taking stock before buying tools.

And those who start early have an easier time. Because waiting is a decision, too. The AI still works with the brand, just with whatever it finds. And "whatever" is the opposite of a brand.

This text was written in dialogue with AI. The machine helped with the research and the wording. People decided what stayed in. And the people accountable for it are named above this text.

Header image: Unsplash photo, adapted using AI.