A brand book used to be a PDF. Then it became a Notion page or a Figma file or some other living document. The next step, which is happening already in some places, is the brand book becoming a JSON file.

What that looks like in practice is a structured data file with everything the brand cares about expressed as fields. Colours as a list of named tokens with hex values and roles. Typography as a set of named scales with weights and line-heights. Spacing as a numeric scale. Logo references as file paths. Tone of voice as a structured object with lists of permitted vocabulary, forbidden vocabulary, sentence-length ranges, punctuation rules. Photography direction as a set of prompt fragments tagged by use case. Motion as easing curves and duration values, expressed as numbers a tool can read.

The reason is multiplication of clients. A brand now has to be readable by a number of different tools at the same time. A designer working in Figma needs to be able to pull the brand's colours and type. A developer writing front-end code needs the same values in their CSS variables. An AI image tool needs photography direction in a format it can parse. A copywriting assistant needs the tone of voice as a system prompt. All of those benefit from a single source of truth they can pull from rather than copying out of a PDF.

Design tokens have been a piece of this for a few years already. Salesforce, IBM, Shopify and others built early tooling around the idea, and the W3C Design Tokens Community Group has been working on a standard format since 2019. The pattern was confined mostly to large product organisations until recently. What's shifting is that smaller brands are starting to need it too, partly because the tools that consume design tokens have got cheaper and more accessible. Figma Variables exports to JSON. Tokens Studio sits on top of that. Style Dictionary turns JSON into CSS, Swift, Android, whatever the downstream needs. The infrastructure is there.

There's something familiar about this if you've looked at the work of Unimark in the 70s, or the systems Vignelli built for American Airlines and Knoll. Those weren't brands expressed as a single image. They were brands expressed as a set of rules. A grid, a typeface, a colour, a paper stock, applied consistently across applications. The deliverable was often a thick book of specifications with a precision that feels closer to engineering documentation than to art direction. JSON is the same impulse, just translated into a syntax that a piece of software can read directly. The Swiss school would probably have been comfortable with it.

The PDF doesn't disappear. It's still the version a human reads. The JSON is the version a tool reads. Both can exist alongside each other, and the better setups are starting to generate one from the other. A change to the source JSON updates the documentation site automatically. The hex code on the colour swatch page matches the hex code in the developer's code because both are reading from the same field. This is mostly invisible to the client but it removes a class of bugs that have lived in brand work for as long as brand work has existed.

There's a small risk in this worth naming. Codifying everything makes it harder to break the rules later. A brand voice expressed as a structured object with permitted-words and forbidden-words lists is a brand voice that gets followed literally by the next tool that reads it. The looseness that lets a writer ignore a rule in service of a better sentence is harder to preserve in a schema. The systems that handle this well are leaving space for it. The systems that don't end up with brands that read as if a robot wrote them, because in a sense one did.Where this is going, at least for smaller businesses, is brand context files being asked for in briefs. A client setting up an AI assistant for their content team wants something to feed it that isn't a thirty-page PDF. A developer building a new product surface wants a file they can import. A marketing person setting up an automated content pipeline wants something machine-readable they can plug in. The format and the request are still informal in most cases, but the direction is clear enough to plan for.

It connects to older practice more than it looks like it should. A brand expressed as a set of rules instead of a single image is not new. It's the dominant approach of mid-century design systems. The syntax is different. The intent is roughly the same.