Are your brand guidelines AI-ready?

Another chance for me to preach about how important quality inputs are

We already know that the same brand input won’t produce the same output twice. Swap the writer, swap the AI tool, swap both — the result drifts. How far it drifts depends on the quality of the input.

When it comes to writing on-brand, I believe “quality” input needs to be redefined.
Most brand guides read like this: “We are confident, approachable, human, and speak to customers like a friend.” This might be fine for an in-person workshop, but it’s useless when someone actually has to write with it when there are no supporting rules or examples.

A senior writer can overcome vague instructions with instinct that’s been built over years working at the same company. But others, including generative AI, can’t.

My team sees this constantly. For example, we recently had negative feedback from a client on some web copy we’d delivered. What the client didn’t realise is that almost every section they didn’t like was copied from their own existing pages as we were optimising – not overhauling – their content.
Off-brand copy doesn’t need AI to slip through. Multiple writers producing content for the same brand, with no concrete rules to check against, will drift the same way a model does.

As the pressure goes up to do more with less, so does generative-AI use

In Deloitte’s 2025 Connected Consumer research, 53% of US survey respondents said they use generative-AI regularly – including 34% who use it for work. With so many different teams involved with content production and customer comms, inputs and outputs can vary – and this can lead to a slip in brand identity that deteriorates fast.

There’s another reason this matters, too. When customers put a query into AI search, the answer will be based on whatever the model can find. And it rewards consistency. This is where things start to fall apart for large organisations who do not have brand guidelines with rules, references and rationale behind them.

My word(s) of the week: brand tokenisation

This is where Sarah Balmer comes in – she recently taught me the phrase brand tokenisation and believes it will be one of the biggest conversations in 2027.

The idea is that you define your brand’s voice and rules in one canonical, structured document (built on rules, references and rationale) so it can plug into whatever AI tool anyone is using, regardless of how much accumulated brand knowledge they personally carry. The goal is to minimise the disparity between 15+ different people’s outputs that continue to drift over time.

Brand tokenisation is not about creating a new brand guide, it is about translating the ideas that already exist.

An important caveat: It doesn’t get you out of maintaining it. Someone still needs to own it. What’s different is that brand tokenisation changes where that update happens: once, centrally, instead of fifteen times, unevenly.

Technical skills aside, any senior marketer can put together the raw material needed to get this result. Just start with the rules, references and rationale.

Examples of contextual rules

Collecting feedback and agreeing on revisions

When stakeholders can’t agree on what “on-brand” actually means, at Avion, we sit down one-on-one with the client to examine copy from another company they don’t own. It can be wildly different, like Nike or Amazon, or a challenger brand fighting its way to the top – the point is that it’s not theirs, so they can be honest without repercussions.

Using another brand’s copy for reference can do more to align people on comments than asking them to please explain.

Anticipate that legal will have the final say

Nothing in this article matters if legal doesn’t like it. So why not get them involved, or at least factor in their point of view, as early on as possible.

By incorporating rules and examples into brand guidelines, legal teams can look at something more concrete than adjectives that are open to interpretation. It also allows you to record a paper trail – which is critical for highly regulated industries.

We used editorial judgement won’t hold up in a compliance review, but Here’s the rule, here’s the reference, and here’s why it was approved might.

A major challenge is that legal will always default to making things vague to minimise risk – the opposite of what marketing teams are trying to do when they want their brand to stand out. This framework doesn’t change that negotiation, but makes it easier to have.

What to do this week

1. Refresh your brand guidelines to use the format: rule, references, rationale.
2. Get brand, legal and compliance in a room together to discuss and agree.
3. Use the approved as custom instructions for every AI tool your team is using.
4. Name one owner. Brand tokenisation makes the rules easy to distribute, but they don’t maintain themselves.

Questions you might ask before getting started

We already built a custom GPT for this. Isn’t that the same thing?

No. A custom GPT is built by one person, for one project, and it dies the moment that project ends, or that person moves on. What you want is a structured document the whole team keeps drawing from, not a personal workaround someone else has to reverse-engineer later.

Do I need to be technical to do this?

No. It’s the same skill as writing a good creative brief — turning vague brand feel into concrete, specific examples. That’s copywriting and editorial judgement; no code involved.

How do I know this is actually worth the time?

Run a before/after test using only adjectives VS the three Rs framework (rules, references, rationale). If you can show your own team a real example of AI outputs that sound inconsistent, the case makes itself. Nobody needs convincing once they’ve seen their own brand butchered by a model.

Isn’t this the same as just training the team to prompt better?

No. Prompting skill is personal and it doesn’t scale. A new starter joins or someone’s having an off day and you’ll see consistency drop. A structured three Rs framework takes that dependency out of any one person’s hands.

The bottom line for marcomms teams

Marketers who sharpen their brand guidelines benefit in two ways: their teams move faster, and AI tools get better at spotting what’s on-brand versus what’s not in a first pass. Today’s deadlines don’t leave room for 100% human-written content, 100% of the time, so the more precise you get about what “on-brand” means, the closer AI gets to matching it without you having to fix it after the fact.