Thinking · Newsletter
Hooks and creative briefs: the complete guide
A hook is a promise. Five hook families, motivation over demographics, where concepts come from, writing with a model without slop, the concept system, and the pre-flight checklist.
TL;DR. A hook is a promise made quickly enough that someone stops to find out whether it is true, not a pattern interrupt. There are five working families of hook (visual, verbal, situational, demonstrative, contrarian) and most libraries are concentrated in one or two, so the cheapest new test is the family you have never tried. A brief that produces testable hooks is written to a motivation rather than a demographic, specifies the opening (more than one), names what varies and what is held constant, and knows where the click will land before anything is shot. Concepts that survive contact with an account come from the comments, the reviews, the existing library decomposed, and the category next door; the whiteboard has the worst hit rate. This is the whole method from first second to signed checklist.
What a hook does
"Your hair is not falling out. It is thinning at the part." That is a promise. Someone who stops on it has a specific question, and that question is now the only thing they care about. The pattern-interrupt framing has done damage because it buys attention the ad cannot spend: something impressive and unrelated stops the thumb and then has nothing to say. Upptic's 2026 framework calls the first two to three seconds "the pattern interruption that earns attention"; we would say the interruption is the mechanism and the promise is the content, and only the promise is testable. See a hook is a promise.
The five families
| Family | What it does | Fails when |
|---|---|---|
| Visual | Something on screen is wrong, surprising or beautiful enough to stop a thumb | It is impressive but unrelated to the product |
| Verbal | A line that opens a loop ("Nobody tells you this about...") | The loop is never closed, which trains people to scroll past the next one |
| Situational | A recognisable moment the viewer has lived | The situation is generic enough to belong to any brand |
| Demonstrative | The product doing the thing, immediately | The thing is not visually interesting, which is most products |
| Contrarian | A claim that argues with what the viewer believes | It is contrarian without being true; the comments will say so |
Classify the live library against these five. Almost every account is heavy in one or two. The gap is the test. Underneath all five sit four things about attention that have survived every platform change: specific beats general, an open loop holds (if you close it), people believe other people, and contrast creates attention rather than volume. A concept that uses none of them is relying on production value, the most expensive and least reliable thing to rely on. See four things about attention.
Motivation, not demographics
The fastest free improvement to any brief: replace the demographic with the motivation. A demographic describes who might buy and says nothing about why, which is the only thing the ad communicates. Motivations are testable and the finding travels; age brackets give you nothing to carry forward. List the five distinct reasons someone buys (reasons, not segments), then count how many the live library covers. Most cover one or two, heavily. That gap is the cheapest creative opportunity available because it needs no better production, only a different brief. See brief to motivation, not to demographics.
Where concepts come from
Four places, in order of hit rate. The comments on your ads and competitors', the only place buyers describe the product unprompted, objections included. The review and support backlog, a list of promises made and not kept, which is a list of promises worth making properly. The existing library decomposed: not "run it again" but "what was the mechanism". The category next door: gaming borrows from commerce and back, and the borrowed device is fresh because the audience has not been trained on it. A concept enters production with a note on which input produced it; blank means whiteboard. Booksy described the same move in October 2026: an intelligence layer that reads what is working before any brief is written. We read the category as well as the account, through Flixr Studio. See where ideas come from and creative operations, defined.
Writing with a model without producing slop
"Write me an ad script for X" returns the average of everything the model has seen, and the average ad is bad. What changes the output: give it the objection, not the product ("write to someone who thinks this is overpriced"); give it a voice sample (three real customer comments); constrain the structure ("first line a question, no product until second five, end on a decision"); and ask for the bad versions and which it would cut. What it still cannot do is know what your buyer says. That comes from the comments and calls, and it is the input that separates a script that could be for anyone from one that could only be for you. See scripting with a model.
The brief that produces testable creative
A concept system, not a film: one approved idea with the variables named before anything is shot. Opening, performer, problem frame, demonstration, ending, length. One idea becomes thirty assets and the account tells you which pocket of the audience each argument finds. There is no perfect ad; the information is in the spread, not the specimen. See there is no perfect ad.
The pre-flight checklist
Before the shoot is booked: the question this batch answers, in one sentence; the variation axis, written down; the opening, specified and more than one; the role, not the demographic; the product confirmed in hand. Before the day: approvals sequenced with named approvers and dates; the compliance line agreed; fallback framings planned (two angles per beat); an asset-count target (if nobody says a number the answer will be four); and the destination known, because a page that changes the subject invalidates the batch after it is made. Signed, not acknowledged. See the pre-flight checklist and the pre-production mistakes that cost most.
The hook's other half
The promise the hook makes has to be kept by the page the click lands on. Most are not: the ads are specific and the pages are generic. Audit it in an afternoon: write each top ad's promise in one sentence, open its destination, write the first sentence a visitor reads, count the mismatches. See landing page vs product page for paid ads.
Questions people ask
What makes a good ad hook? A specific promise made in the first second or two that the rest of the ad keeps. Visual, verbal, situational, demonstrative and contrarian are the five working families.
How do I write a creative brief for performance ads? To a motivation rather than a demographic, with the opening specified more than one way, the variables named, a compliance line agreed, an asset count set and the destination known.
How long should a hook be? Long enough to make the promise. On social, under two seconds. On AppLovin and other fullscreen placements the hook still decides whether the viewer watches or waits, but the ad has 30 to 60 seconds to keep the promise.
Can AI write ad scripts? It can draft well when constrained with the objection, a voice sample and a structure. It cannot supply what your buyers actually say.
Sources: Flixr newsletter pieces, April to September 2026; Upptic, "Best Ad Creative Strategy for 2026"; Gamesforum, "Inside Booksy's Creative Playbook", October 2026.