How long should a creative brief be?

Long enough to answer the four core questions and no longer — keep it to a single page or screen. A brief that runs to three or four pages is usually a sign the objective was never actually singular; multiple objectives got bundled into one document instead of split into separate assignments, and the maker is left guessing which one to prioritize when they conflict.

Who should write the brief — the person commissioning the work or the person doing it?

The commissioner writes it, because the brief encodes decisions — objective, audience, claims — that belong to whoever owns the outcome, not whoever executes it. A maker who receives a gap in the brief should ask rather than fill it in silently; a maker who writes their own brief from scratch is effectively making the commissioner's decisions for them, which removes the accountability a brief exists to create.

What's the difference between a brief and a brand style guide?

A style guide is persistent — voice, visual rules, approved claims — and applies across every asset a company produces until it is deliberately revised. A brief is disposable: written for one asset, answered once that asset ships, then archived. A style guide tells a maker how the company always sounds; a brief tells a maker what this one piece has to do. Assume the style guide rather than restating it, and spend the brief's limited length on what is specific to this one assignment.

Does a creative brief still apply when the maker is an AI writing system rather than a person?

The same four questions apply, and two sections get added that a human maker rarely needs written down. The first is an anti-echo list: the pages you already publish next to this one, named, with an instruction not to restate them — two systems briefed on adjacent topics will otherwise produce the same explanation twice, and a library built that way drifts toward one article republished under several titles. The second is a closed source whitelist: the named places a figure may come from, with everything outside the list barred however plausible it looks. Both sit in how to brief AI writers as pass-or-fail laws with stated grounds for rejection, which is what separates them from guidance. How much else has to be stated outright is a question about the systems in use at the time, and the an

What is a creative brief in marketing

Guide · Glossary & Definitions · 5 min read · last verified 2026-07-29

Reviewed before publication Editorial board Independent commercial review

In shortA creative brief is the artifact that tells a maker what to achieve, for whom, and with what evidence. What it must answer, what it refuses, and why agencies keep asking for one.

A creative brief is the short document that tells a maker — a writer, a designer, an agency, an AI system — what an asset must achieve, for whom, and with what evidence, before anyone starts producing anything. It exists to move decisions to the front of the process, where they are cheap to change, instead of leaving them to surface as opinions during review, where they are expensive to change. The mechanism is that simple: a decision made before anyone builds anything costs a conversation, and the same decision made after a draft exists costs the draft. Length is not what separates a brief that works from one that does not — a document that answers the four questions below is doing the job whatever its length, and one that only records preferences is not doing it at any length either.

What a brief must answer before work starts

Every functional brief answers the same four questions, regardless of format or house template. What is this asset supposed to change in the audience's head or hands — an objective a stranger could check for pass or fail, not a mood. Who is reading, watching, or hearing it — specific enough that a writer could picture one person rather than a market segment. What already exists that this must not contradict — the claims, proof points, and constraints inherited from a messaging hierarchy or a prior campaign. And what evidence backs any claim the asset intends to make — a source, not an assumption borrowed from a competitor's website or a guess at what buyers want to hear.

Each of the four is on the list because of the specific failure that follows from leaving it blank. An objective without a way to check it produces work that only taste can evaluate, and taste does not scale across a team. An audience of everyone produces copy calibrated to no one in particular. Unstated inherited claims produce an asset that contradicts the last one, and someone then has to reconcile the two in public. Claims with no source produce assets that legal, a fact-checker, or an unhappy customer will eventually strip out, at a cost higher than checking the source would have been.

What a good brief refuses

The useful brief is defined as much by what it declines to accept as by what it specifies. It refuses an audience of everyone, because a brief trying to reach everyone gives the maker no way to choose a tone, an example, or a level of detail — the maker will choose anyway, just without guidance. It refuses an objective that cannot be measured or observed, because unmeasurable objectives cannot be argued with later, which means every revision request becomes a matter of opinion instead of a return to the brief. And it refuses claims without sources: a line like "buyers care most about X," or a statistic lifted from a competitor's landing page, has no place in a brief unless someone can point to where it came from. A brief that asks for an AI-visibility angle needs the same kind of sourcing as any other claim it makes — a scan that shows which sources currently get cited for the buyer question in play, the kind of evidence a tool like Magrios produces, is something a brief can cite, where a guess is not.

The shape of one, in practice

Templates vary by team, but a functional brief commonly takes something like this shape, compact enough to fit on one screen:

Objective: what changes, for whom, checkable after publication

Audience: one specific reader, not a segment

Single most important thing: the one claim if nothing else lands

Proof: sources for every claim, named and linked

Format and length: the container the maker is building for

What this is not: the boundary that keeps scope from drifting

Deadline and reviewer: who signs off, and by when

The last line matters more than its brevity suggests. A brief without a named reviewer invites a review cycle where everyone with access to the document feels entitled to add a preference, and preferences accumulate faster than objectives do. That core shape holds whether the maker is a human writer, an agency, or an AI writing system — what changes across those makers is how much has to be spelled out rather than inferred, which with the systems in use today is more than a house template usually carries.

Why agencies keep asking for it

An agency that keeps requesting a brief before starting work is not stalling — it is declining to guess on the client's behalf at the client's expense. Unbriefed work forces an agency to invent an objective, an audience, and a set of claims, then defend those inventions during review as though they were the client's own decisions. That defense burns hours the client is paying for, and what the agency defends is a target it constructed rather than one it was given — which crowds out the question worth asking, whether the work does the job. A brief shifts that cost to the front of the engagement, where it is a single short document instead of two rounds of revisions built on a guess. Briefing an agency on AEO work specifically narrows further — buyer questions, a client-owned measure, and scope — but the underlying discipline is the one described here.

What it feeds and what it draws from

A brief is a translation layer, not a source of its own. Upstream, it draws its objective and its claims from whatever the company has already decided — a positioning document, a messaging hierarchy, a prior scan or piece of research. A brief invented from nothing, with no upstream document to translate, generally gets filled in with the writer's best guess about what the company would probably want, and best guesses vary by writer.

Downstream, the brief becomes the standard the finished asset is checked against. QA before an asset ships is, in large part, a comparison back to the brief's objective and claims — did the asset do the one thing it was supposed to do, and can every claim in it be traced to a source. Assets built without a brief have nothing to check against except opinion, which is why they take longer to approve and longer still to defend once published. The brief treated as disposable paperwork before work starts is usually the same document everyone wishes they had kept once the work is finished and someone asks where a claim came from. Keeping approved briefs alongside the assets they produced, in an organized asset library rather than scattered across chat threads, is what makes that lookup possible six months later.