How to write brand rules an AI can check

Most brand books are written to inspire. They open with a mission statement, a mood board and a paragraph about the brand’s personality. That is useful for a designer starting a new campaign. It is not much use when someone asks a simple question: does this banner follow the guidelines?

To answer that question, whether a person or an AI is doing the checking, you need rules that are specific enough to test. This guide shows how to turn a brand book into a list of checkable rules, with a sample table you can copy.

Step 1: Pull out the rules that are already checkable

Start by reading your brand book and separating the concrete rules from everything else. A rule is checkable if two people looking at the same creative would agree on whether it passes.

Use this test on each line of the guidelines:

• Checkable: “The logo must not be placed on a photo without the white backing shape.”
• Not yet checkable: “The logo should always feel confident and clear.”
• Checkable: “Every offer must include the line: Terms and conditions apply.”
• Not yet checkable: “Make sure legal is covered.”

Copy the checkable rules into a new document. For the rest, ask what the guideline is really trying to prevent, and whether you can write that as a concrete rule. Some lines (the brand story, the mission) will never be checkable, and that is fine. They stay in the brand book. They just do not go on the rules list.

Step 2: Write colours as hex codes

Every approved colour should be written as a hex code, not a name or a description. “Our signature blue” means nothing to a checker. “#1F4FD8” does.

For each colour, write down:

• The name your team uses, for reference.
• The hex code for screen work.
• Where it is allowed: backgrounds, text, buttons, accents only.
• Any combinations that are not allowed, such as a light colour for text on a white background.

Be explicit about what is not on the list. A rule like “use only the colours listed below for backgrounds and text; photography is exempt” gives a checker a clear boundary. Without the exemption, every photo would get flagged.

Step 3: Name fonts exactly

Fonts should be written by exact family name and allowed weights. “A clean sans serif” is a description. “Inter, in Regular, Medium and Bold only” is a rule.

Note the cases where a fallback is acceptable, such as a system font in email or on a platform that does not support custom fonts. If you do not write these exceptions down, someone will flag them as errors every time, and people will start ignoring the check.

Step 4: List required legal lines word for word

Required legal lines should be written out exactly as they must appear, with a note on where and when they apply. This is the area where a precise rule saves the most trouble, because a missing disclaimer is the kind of error that gets noticed after launch.

For each line, write:

1. The exact text, including punctuation.
2. Which work it applies to: all ads, only offers, only a certain product, only a certain region.
3. Where it must appear, if that matters, for example on the final frame of a video.
4. Whether any variation is allowed, such as a shorter version on small formats.

If your legal team has different lines for different markets, list them separately and say which applies where. A checker cannot guess which market a banner is for unless the project tells it.

Step 5: Build a banned words list

A banned words list is a plain list of words and phrases that must never appear, ideally with a preferred alternative for each. It is one of the simplest rules to write and one of the easiest to check.

Typical entries include:

• Claims your legal team will not allow, like “guaranteed” or “the best”.
• Old product names that have been retired.
• Competitor names, if your policy is never to mention them.
• Words that clash with your tone, like “cheap” for a premium brand.

Adding the alternative turns a flag into a fix. “Do not use cheap; use affordable or great value” tells the copywriter exactly what to do next.

Step 6: Describe tone with examples

Tone rules work when they come with examples. Adjectives like “friendly”, “confident” and “human” are open to interpretation. Pairs of example sentences are not.

For each tone trait, write a one-line description and then two pairs of examples:

Friendly, not casual. Write like a helpful person, not a mate at the pub.

Write: “Your order is on its way. Here is how to track it.”

Avoid: “Woohoo!! Your stuff is coming, yay!”

Tone is the area where any checker, human or AI, needs the most judgment. Treat tone rules as prompts for a closer look rather than hard failures. If you want reviewers to give useful tone feedback too, our guide to clear design feedback covers how to phrase it.

Step 7: Give each rule a severity and test it

Every rule needs a severity, so the person reviewing knows what to fix first. Then test the full set on work you already know is right or wrong, before you trust it on new work.

Here is a sample rules table you can adapt. The values are examples, not real brand rules:

To test it, pick a handful of past creatives: a few that were approved and a few that were sent back for brand issues. Run your rules against them, by hand or with a tool. If a rule flags good work, it is too strict or missing an exception. If it misses a known error, it is too vague. Rewrite and run it again.

Once the rules hold up, decide where they run. In Yesquill, admins write these rules once and the AI brand check reviews each upload, pinning issues on the creative with a severity before any human reviewer sees it. It can run automatically on every upload, and it is available on the Pro and Agency plans (details on our pricing page). You can also use the same table as a manual checklist alongside our pre-send review checklist.

Either way, keep the rules list alive. When a new legal line arrives or a product is renamed, update the list the same day. A short list that is always current is worth far more than a thick brand book nobody opens.