Every Ad Format Has a Load-Bearing Wall

Scroll any consumer brand's ad account for ten minutes and you'll see the same shapes repeating across every category. The comparison table with your product's column winning. The fake breaking-news chyron. The iPhone note. The text-message thread. The review card with the stars. The whiteboard explainer. These are ad formats, and they're the folk vocabulary of consumer advertising, passed around in swipe files and media-buyer group chats, named informally, understood by everyone and examined by almost no one.

I've bought a lot of ads across the consumer brands I've operated, and it took me embarrassingly long to see that a format isn't a look. It's a piece of persuasion machinery with some parts that carry the load and some parts that are just paint. Confuse the two and the format stops working, usually without telling you why.

What's fixed and what's free

We went through this formally at Jinn — cataloguing the recurring ad formats in consumer advertising, forty-two of them, for our creative product. The most useful output of that work was a distinction we ended up writing into every entry: for each format, what is fixed when you adapt it to a new brand, and what is free.

Take the breaking-news format. Free: the product, the image, the palette, how playfully you wear the frame. Fixed: the news-ticker grammar (the "BREAKING" plus the urgency trimmings plus the timestamp) and a headline written as news, not ad copy. Our catalog entry puts it in five words: "new from us" kills it. The format works because years of real news alerts wired an attend-now reflex into everyone's thumb, and the reflex fires on the grammar of news. Write the headline like an announcement from a brand and the grammar collapses; you've paid for a news costume and delivered a press release.

Or the whiteboard explainer. Free: the topic, the style. Fixed: visible hand-drawn imperfection, and one idea built incrementally in front of the viewer. Polish it into an infographic and you lose the classroom trust that was the entire mechanism: the drawn diagram reads as explanation rather than pitch, and watching the idea assemble feels like understanding it.

Or the us-vs-them comparison table. Free: the visual form — table, split photo, receipt gag. Fixed: axes your brand genuinely wins and that buyers actually weigh. A rigged-looking table backfires into distrust, because the reason the format converts is that the reader feels they ran the comparison themselves. A comparison that's obviously staged returns ownership of the conclusion to you, and the reader discards it accordingly.

The text-message format is the one I'd hand a new media buyer first, because its wall is the hardest to see. Free: the relationship, the product, the joke. Fixed: the recommendation has to sound like a friend actually talking, off-hand, mid-topic, a little teasing. The moment it turns into a benefits recital with a contact name on top, the reader files it under ad, and the whole point of borrowing a private conversation is gone. The screen has to look exactly like a real phone for the same reason. A friend's text is the highest-trust recommendation channel humans have, and the format works by counterfeiting the setting so the endorsement inherits its warmth. Get the screen right and the voice wrong and you've built a very convincing envelope around a flyer.

The emergency format has a wall that's really a fork. It runs in two modes, straight urgency and open joke, and both work. The straight version needs a true deadline or a real stock count. The joke version has to be unmistakably a joke. What kills it is the half-serious middle, the ad that wants urgency's reflex without committing to either the fact or the gag. Emergency visual language makes the body answer before the mind has read anything; catch it bluffing once and the body stops answering, for your next ad too.

That's the pattern across all forty-two. Every format has a load-bearing wall. The brands that use formats well renovate everything except the wall. The brands that use them badly knock the wall out and keep the wallpaper.

Some load-bearing walls are made of facts

In a subset of formats, the load-bearing element is a factual claim — and that stops this being craft advice and makes it a liability question.

The review-card format works because a neutral third party did the counting; the stars, the aggregate score, and the review count are that party's numbers, not yours. Which means the reviews have to exist. A fabricated aggregate is the exact fraud the format's credibility is borrowed from.

The hard-offer format works off a reference point: the crossed-out price sets what the product is worth, and the gap reads as money lost by not acting. Which means the discount has to be real. The invented "was" price is the classic dark pattern of the category, old enough to have case law.

The comparison format, when it names a competitor rather than a generic "them," is making public claims about someone else's product — a different kind of exposure entirely.

The stat headline is the purest case: one number as the whole ad, everything else subordinate. An exact number is processed as evidence where an adjective is processed as advertising, which means the number has to survive a reader who checks. If it falls, the whole ad was a lie, not a fudge. The before-and-after runs on the same logic; the honesty of the pair is the persuasion, and a doctored before is the failure the platforms police hardest.

When we built our catalog, we attached checks to this class of format — a fabricated-social-proof check, a claims check, a competitor check — on every entry whose mechanism depends on a fact being true. The checks are written into each entry rather than kept in a legal document nobody reads, so the software knows which formats can't go out without a real number behind them, and refuses to fill the slot itself.

In practice that's three different behaviours, because the three kinds of fact fail differently.

For the formats whose mechanism is a real number, the check runs before a single render is paid for. An offer needs its real terms on file. A low-stock alert needs a stock count or a date somebody vouched for. A before-and-after needs the genuine pair; a stat headline needs the stat and its source. If the brand hasn't attached one, the format refuses to generate. It doesn't produce a draft with a placeholder number and hope someone remembers to swap it. The refusal comes before the money, which is the only place a refusal is cheap.

There's a subtlety in there we argued about. The myth-versus-fact format needs a correction that's true and defensible, and a correction is usually not a measured number. So we don't let an attached statistic satisfy that check, even though it would have been easier to wire. A real number about one thing would otherwise license a fabricated "fact" about another, and a check that accepts the wrong kind of proof is worse than no check. It produces a paper trail that says the lie was verified.

For the social-proof formats, the review card, the text thread, the Reddit post, the screenshot of an email, the render can be produced, but it carries a permanent marker saying the social proof is a placeholder, and nobody can export it, our own team included, until a human has looked at it and acknowledged what it is. The text thread is the reason that step exists: of everything in the catalog, it's the one that reads most real.

For the comparison formats, the competitor check has to run on every render. If it isn't available, the format won't generate at all rather than go out unchecked. We'd rather lose a render than ship a named rival nobody looked at.

And this is the honest case against the prompt box. A prompt box has no concept of a load-bearing wall. Ask a general-purpose image tool for a "Trustpilot-style ad" and it will draw you stars, a score, and a review count with perfect confidence, because to the model those are pixels in a style, not claims with a truth value. The tool isn't malicious. It just doesn't know which part of the format is paint and which part is a fact you'll be accountable for.

A catalog is a claim about validation, so I'll state ours honestly

At Jinn this catalog lives inside Vermeer, our creative product, as a working reference the software reads: each format's definition, its fixed-versus-free anatomy, the psychological principle it runs on, the checks it needs, and the customer goal it serves — because the honest way to pick a format is by goal ("handle the objections blocking a purchase," "make cold traffic stop") rather than by which template looks nice.

Every entry in the catalog starts switched off, and formats are switched on for customers one at a time, each after a human has looked at what it actually produces. We built it that way on purpose. A catalog is a promise that each entry works as described, and we'd rather turn formats on slowly than let the list imply validation that hasn't happened.

You don't need our software to use the idea. Take the five formats your brand actually runs and write four lines for each: what's fixed, what's free, why it works on a human, and what would make it a lie. That last line is the one nobody writes down, and it's the one that decides whether the format is compounding your brand's credibility or spending it.

The worksheet filled in for the text-message format, so you can see why the fourth line earns its place:

Notice the fourth line didn't say don't run it. It said know what you're running. Under half the entries in our catalog carry a check at all; the rest are pure craft. But the checked ones are the formats brands reach for when performance dips, because urgency and social proof are what move warm traffic. That's the moment the temptation to knock out the wall is strongest, which is why the line gets written before the dip, not during it.

Forty-two formats went into our catalog. The list was the easy part. Knowing what breaks each one is the product.

Questions people ask

Vermeer is live and invite-first; self-serve signup is off while onboarding is managed by invitation. Formats are switched on for customers individually, after human validation.