More and more marketing is produced by AI as well as by people. Without a shared source of truth, quality slips and distinctiveness fades. A brand code fixes that at the start, not the end.
More and more businesses now have a marketing team that isn't only people. A freelancer writes the case study. An AI tool turns it into five social posts. Another rewrites the product descriptions. A creator makes the video, and an agency uses AI to produce the first round of creative.
Each of them is working from a version of your brand. Ask which version, and you'll often get a shrug.
In the agencies I ran, keeping a brand consistent was someone's job. A creative director or an account director owned it, usually with brand guidelines to work from. It wasn't infallible, but it was always clear whose call it was.
Many businesses don't have that. There's no creative director, no design lead, and often no guidelines worth the name. At the same time, the work has multiplied: more channels, more formats, and dozens or hundreds of outputs a month, made by people and machines. Teams are reaching for AI to keep up, and the question of who holds the brand together often goes unasked.
Left to their defaults, AI tools tend to produce the most typical answer to a prompt. That's useful for a lot of jobs. For a brand, it's close to the opposite of what you want.
The result is what people have started calling AI slop. It isn't usually wrong. It's competent, fluent and interchangeable. It could have come from you or from any of your competitors, and that's the problem.
The volume makes it worse. One off-brand post is a small thing. Hundreds of slightly off-brand outputs a month, each one fine on its own, add up to a brand that slowly stops sounding like itself.
There's a cost to sameness beyond looking like everyone else. The Creator Effectiveness Playbook, from System1, WPP Media and TikTok for Business, looked at what makes creator ads work. Showmanship pulled attention, emotion and brand memory up together: stories, characters, recognisable places and spontaneous expression. The biggest gains came from doing something unexpected.
Straight salesmanship had the opposite effect. Facts and figures, direct-to-camera selling and lots of fast cuts were linked to more skipping, less emotion and lower brand memory.
The study is about creator ads, not AI. But the lesson carries across. The safe, typical version gets skipped, and the safe, typical version is what AI tends to produce when nobody tells it otherwise.
Distinctiveness isn't decoration. Jenni Romaniuk's work at the Ehrenberg-Bass Institute shows that a brand becomes easy to recognise and easy to remember when the same assets turn up consistently: the same colours, phrases, shapes and voice, over years. Every inconsistent use weakens that link a little. (More on building distinctive brand assets on an SME budget.)
That isn't a case for doing the same thing every time. It's a case for fixing the few things that must stay the same, so everything else is free to surprise people. The assets hold steady. The ideas don't have to.
AI can multiply either outcome. Fed the right material, it can apply your assets more consistently than any team could. Left to guess, it erodes them faster than any team could. The difference is what it's working from.
Harvard Business Review recently set out the idea of a brand code, in a piece by Michelle Taite, John Winsor and Will Fernandez. In short, it's a machine-readable record of your brand strategy, your customer understanding and your business rules, which people and AI can both act on.
The contrast with brand guidelines is the useful part.
Guidelines tell a designer how the brand looks. A brand code tells everyone who makes your marketing - your team, your freelancers, your agencies, your creators and your AI tools - what the brand says, how it says it and what it never does.
I've built brand codes with clients over the past few months, and in my experience four sections do most of the work of keeping quality up.
1. Language, written specifically enough to follow. "Friendly but professional" isn't an instruction anyone can act on, and AI certainly can't. The words you use, the words you never use, how long your sentences run, whether you say "customers" or "clients": each of these can be written down and checked.
2. Your distinctive assets, and how each one is used. What they are, where they appear, and how firmly established each one is. A colour you've used for ten years needs protecting. A phrase you started using last month is still being built, so it needs using consistently, not tinkering with.
3. Hard stops and review gates. Some things never go out, whoever approves them: a claim you can't support, a comparison you've ruled out, a regulated phrase. Others need a named person to see them first. Write both down, and make sure every AI tool you use is given them every time.
4. A quality bar someone else can check. "Is it on brand?" can't be tested. "Every figure has a source" can. So can "no sentence over 25 words" and "no claim about competitors". If the person checking didn't write the piece and can still say yes or no, the bar is doing its job.
Most of the damage AI does to a brand comes from two habits: stating a guess as fact, and filling gaps with whatever sounds plausible.
The first rule tackles the guessing. Every entry in the code says how sure you are of it: decided, backed by evidence, a working hypothesis, an assumption, or not yet known. AI is told plainly that hypotheses and assumptions may never be presented as fact. It sounds like admin. In practice, it's what stops a hopeful guess in a strategy session turning up three months later as a headline claim.
The second rule tackles the gaps. When something falls outside the rules, the instruction is to use the closest rule, flag it, and log it for a person to decide. Never invent a new rule quietly. A gap that gets flagged costs a minute. A gap that gets papered over can cost a lot more.
A brand code doesn't take judgement out of marketing. It puts judgement in the one place that matters most. People decide what goes in the code, settle the hard calls and approve what carries real risk. AI works from it every time, at a speed and consistency no team can match.
It works the same way for the freelancers, agencies and creators who make content for you. Briefed from the brand code, they know which assets to use and what's off limits - and still have all the room they need for the showmanship that makes the work worth watching.
The judgement stays human. The consistency comes from the system.
You don't need a large budget or a twelve-week project. You need a few decisions written down, and someone who owns them.
Your brand is now briefed by people and machines, many times a day, whether anyone planned it that way or not. With a creative director in the room, that can be managed by hand. Without one, the honest answer to "briefed from what?" is often a PDF, some habits and a few people's memories.
With AI in the team, that increasingly isn't enough. A brand code writes the brand down once, properly, where everyone who makes your marketing can read it, so the thousandth output sounds like the first.
The brand code sits at the heart of Vector, the six-stage growth system I use with clients at Bonobo & Co, my growth consultancy. If your brand is starting to sound like everyone else's, get in touch.
