The on-brand content checklist: 12 checks before you publish

Most on-brand checklists are made of adjectives. Is it consistent? Is it authentic? Does it feel like you? You can answer yes to all three about a post that is quietly off, because nothing in the question tells you what off would look like. So the checklist gets run, everything passes, and it stops being run.

A check is only useful when it names the failure it is hunting. Then it has an answer instead of a mood, and the whole pass takes ninety seconds rather than however long you can stand to stare at a draft before deciding you feel fine about it.

These are the twelve we use. They run in three passes, in this order, because a verbal problem makes the visual pass pointless and a system problem can invalidate both.

The verbal pass

1. Does this argue a position you would defend if somebody disagreed underneath it? The failure mode is the agreeable nothing: a post saying that consistency matters, or that founders should be authentic, phrased well enough to read as insight. Nobody will argue with it because there is nothing there to argue with. The quickest test is to write the opposite sentence. If the opposite is obviously stupid, you have not said anything yet.

2. With your name and logo removed, would somebody who knows you still pick this out as yours? The failure mode is competent marketing: correct grammar, sensible structure, the median rhythm of professional internet writing, and no fingerprint. This is the most common way AI-assisted content goes wrong, and it never looks like a mistake. It looks like a slightly duller version of you, which is why it survives review. The fix is upstream of the draft, in a voice definition specific enough to write against, and we set out that framework in how to define your brand voice.

3. Can you stand behind every factual claim in it, from memory, without going to check? The failure mode is the borrowed number: a statistic half remembered from a conference slide, a "studies show", a percentage that arrived with the draft and has no source attached. If you cannot say where a figure came from, delete the figure. The sentence is almost always stronger without it, and one unsourceable number costs more trust than it was ever going to buy.

4. Does the opening line promise what the body actually delivers? The failure mode is the grafted hook: "I made a mistake that cost me three clients" sitting on top of a general list of tips. Readers feel the switch even when they cannot name it, because a borrowed hook and your own writing are in different registers and the seam is audible by the second sentence. If the hook promises what the body cannot pay off, cut the hook.

The visual pass

5. Is every colour on this exactly from your palette, or is one of them merely close? The failure mode is the near miss: a background nudged a shade warmer, an accent eyedropped from a photo because the kit was in another tab. Nobody notices one. Six posts later the feed carries eleven brand colours and nobody can say which is the real one. Exact values, checked against a single source, is the only version of this that works, and holding that single source is the entire job of a brand kit.

6. Can a reader tell what is headline, what is support and what is detail within half a second? The failure mode is flat hierarchy: everything in one weight, or three sizes so close together that they read as one grey block. Hierarchy is a ratio problem rather than a taste problem, so if two levels sit within a couple of points of each other, one of them is not a level.

7. Does the logo have its clear space, and was it placed rather than fitted? The failure mode is the leftover logo: shrunk into whatever gap the layout left over, touching an edge, or sitting on the busiest part of an image. Clear space is a rule with a number in it, which makes it one of the very few visual checks with a genuinely right answer. Use the number.

8. Does it still read at thumbnail size? The failure mode is body copy that turns into grey texture at 120 pixels wide. Shrink it on your phone, or look at it from across the desk. If the first thing a scrolling reader can resolve is not the thing you wanted them to read, the composition is wrong, however good it looks at full size.

The system pass

9. Does this agree with what you published last week? The failure mode is the invisible contradiction: two posts written eight days apart in different moods. Tuesday says pick a niche and hold it, next Monday says stay broad while you are learning. Both are defensible on their own. Together they read as somebody who has not decided, and the reader who saw both is the reader who was paying most attention.

10. Does it agree with your other formats? The failure mode is format drift: the carousel uses the current palette, the stories still carry the pre-refresh one, and the static posts have picked up a fourth typeface from a template. Formats drift apart because they get made in different places on different days, which is most of the time. The structural fix is that every format is generated from the same profile rather than assembled independently.

11. Is the description this was generated from still current? The failure mode is silent staleness: you changed what you sell in June, the profile still describes the March business, and everything made since has been faithfully on brand for a brand you no longer run. This is the most expensive item on the list because it is invisible from the output side. The post looks fine. It is answering an old question. It is also the source of brand drift that people miss most reliably.

12. Is the format carrying the idea, or is the idea being stretched to fill the format? The failure mode is the ten-slide carousel built out of a four-slide thought, with six slides of restatement in the middle. Readers leave at slide five, which your analytics will eventually tell you and your draft never will. If the idea is one paragraph, publish one paragraph.

What is worth automating, and what is not

Roughly half of this list is machine-checkable and half is not. Pretending otherwise is how a checklist turns into theatre.

Genuinely automatable: anything with a number in it. Palette values are exact strings, so a mismatch is a comparison rather than an opinion. Clear space is geometry. Type hierarchy is a ratio between sizes. And whether the source description is thin, vague or out of date is a property of the description, not of any individual post, so it can be assessed once and reused across everything generated from it. That last one is what our Quality Engine is for: scoring the profile per component, and tracking staleness so that changing a source field marks the work that depended on it instead of leaving you to remember.

Partly automatable: the ones with a measurable floor and no measurable ceiling. A tool can enforce a minimum type size and flag a weak contrast pairing, which catches most thumbnail failures, but it cannot tell you whether the composition earns a stop. Agreement across formats is structural rather than judged: if every format reads one profile, the drift mechanism is removed, though nothing has assessed whether the result was any good.

Irreducibly yours: everything in the verbal pass. Whether the claim is true. Whether you would defend the position. Whether the sentence sounds like you rather than like something in your general vicinity. Whether this needed publishing at all. No scoring system reaches those, and one that claimed to would be worse than having none, because you would believe it.

Automate the checks that have a right answer. Keep the ones that require you to hold an opinion.

Running it without turning it into ceremony

Run all twelve for the first month, and keep a note of which ones actually catch things. It will be two or three, and they will be yours specifically. Some people never miss a palette value and contradict themselves monthly. Plenty of people are the exact reverse.

After that, run your two or three every time, and the full twelve whenever something has changed: a new offer, a new audience, a new format, or a batch made in a hurry. Hurry is when this list earns its keep, because hurry is when the near-miss colour and the borrowed statistic get waved through.

The reason we moved the checkable half into the product rather than leaving all twelve as a document is that a document asks you to be diligent at precisely the moment your diligence has run out. What Repic does today, and the features overview is the honest list, is hold one profile, generate the formats from it, and score that profile so the system pass is not a memory exercise. The verbal pass stays with you, and we would not want it any other way.

Repic is in early access and invitations go out in signup order, so if you would like to run this against your own work, you can request access.