The first two seconds decide most of a reel

Open Instagram and watch your own thumb for thirty seconds. You will stop on two or three things and skip perhaps forty. You did not make forty decisions. Something in the first moment of each one made the decision for you, and on the reels you posted last week, that same moment made it for everybody else.

Most advice about hooks tells you to write a better one and stops, which helps about as much as being told to write well. A hook is a small piece of craft with a limited number of shapes, and the shapes have names. Once you can name them you can tell that you have used the same one four times this month, and you can tell when the one you reached for stopped working two years ago.

What follows is that catalogue: the five places a hook can live, what each opening promises, which ones have been used so heavily that they now read as a signal to scroll, a way to test yours before it costs you a post, and the things a hook has no power over at all.

What a thumb does before you have said a word

A reel is not competing with other reels in some general sense. It is competing with the next reel, which is a quarter of a second away and costs nothing to reach. That turns the first moment into a filter rather than an introduction. The viewer is deciding whether to stay, and leaving is free.

The speed of that decision is not folklore. A Facebook-commissioned Nielsen study, reported by Marketing Dive in 2015, found that people who watched less than three seconds of a video ad still showed a 47% lift in ad recall. The encouraging half of that finding is that video registers on people very fast. The discouraging half is the same fact from the other side: three seconds is long enough for somebody to take everything they are ever going to take from your reel, and go.

Then there is what the app does with the decision. Instagram's own explainer, published in May 2023 as "Instagram Ranking Explained", says the reels system predicts how likely you are to watch a reel all the way through and how likely you are to reshare it. In January 2025 Adam Mosseri, who runs Instagram, named the three signals that matter most as watch time, likes and sends, and told creators to look at average watch time, likes per reach and sends per reach in their own insights.

Reach compounds all of it. TechCrunch reported in April 2024 that Instagram had changed how it ranks recommendations: eligible content is shown to a small audience first, the reels that audience engages with go to a slightly wider one, and the best of those go wider again. A reel that loses its first small group never meets the second. The shot list, the edit, the caption you rewrote four times: all of it sits behind a gate that a few hundred strangers either open or do not.

That is why two seconds is a useful working rule even though no platform publishes such a number. Read it as the size of the promise you have to make before anybody agrees to hear the rest, rather than as a countdown you are racing.

A hook lives in five places, and most people write one

Ask somebody to improve their hook and they will rewrite the first sentence they say out loud. That sentence is one of five surfaces, and on plenty of posts it is not the one doing the work.

1. 01The first frame. Seen before anything is heard, and on a muted phone in a waiting room it may be all there is. A frame showing a person already mid-action reads instantly. A frame showing a title card reads as an advert, and gets treated as one.
2. 02What you say first. The spoken line, and the one everybody rewrites. It has to land inside roughly the first three seconds and it has to be a sentence rather than a greeting. "Hi guys, so today I wanted to talk about" has spent the entire hook on nothing.
3. 03The words burnt into the frame. The surface most often left empty, and the one with the clearest evidence behind it. In an April 2019 survey of 5,616 US adults by Verizon Media and Publicis Media, 80% said they were more likely to watch a whole video when captions were available, and 69% said they watch video with the sound off in public places.
4. 04The first line of the caption. Instagram truncates a caption, so you get roughly one line before somebody has to tap. On a photo or a carousel that line is the hook. On a reel it is the second chance for the person who watched it silently and could not follow what was going on.
5. 05The first carousel slide. A carousel gets exactly one frame to earn a swipe. Slide one is the whole pitch rather than a cover, and a slide that says "Part 1" has pitched nothing at all.

Write all five for one post and you will notice they start to disagree with each other, which is the useful part. If your on-image text is a summary of your spoken line, one of the two is wasted. If the first line of your caption repeats the words already on the frame, you have written one hook three times and covered a single angle when you could have covered three.

proof: the five surfaces are a closed list in the analysis: spoken, first frame, on image, caption, carousel slide (server/intelligence/taxonomy.ts)

There is one more reason to write all five. A reel that gets resurfaced weeks later, in the reels tab, to somebody who has never seen your face, arrives without any of the context you had in your head when you shot it. The surfaces are what travels with it.

The openings, and the ones that have gone tired

Here is the catalogue. Every opening in it works and every one of them has been worn thin somewhere, which is what the fourth column is for. The point of naming them is being able to look at your own draft and say which one you used.

Two things fall out of that table. The first is that weak hooks tend to be weak in the same way: a promise gets made and never gets paid, and the viewer learned that pattern from the forty reels they saw before yours. The second is that tiredness is local. "I did this for 30 days" is exhausted in fitness and completely fresh in, say, learning to cook one dish properly.

Notice also that none of these openings require you to be an interesting person. They require you to have something specific to say and to say the sharpest part of it first. The most common mistake in a hook is saving the good bit for the middle, where it belongs in an essay and nowhere else.

proof: the vocabulary runs to 22 named hook types and it deliberately includes the weak openers, so the analysis can tell you yours is one (server/intelligence/taxonomy.ts)

How to check a hook before you post it

Reading your own hook back is unreliable, because you already know what the post is about and you cannot un-know it. So the check has to be mechanical, and it has to be run on the draft rather than remembered as a principle.

The last one catches more bad hooks than the rest put together. Openings accumulate throat-clearing without anybody deciding to add it: a logo, a breath, a greeting, a shot of you walking towards the camera and sitting down. Cutting the first two seconds is a test you can run on anything you have ever filmed, and the answer is usually to keep the cut.

There is also a real test, on real strangers. Instagram's trial reels, which TechCrunch reported rolling out in December 2024, show a reel to accounts that do not follow you while hiding it from the ones that do. Instagram gives you views, likes, comments and shares after about 24 hours and compares that trial against your previous trials. A test with a control group is the only kind worth much, and this is the first one the app has handed to small accounts.

What a hook cannot fix

A hook buys the next three seconds. It cannot buy the next thirty, and it certainly cannot make somebody keep a post they did not want in the first place.

The clearest way to see this is a post that reaches a great many people and gets kept by almost none of them. In the demo account this site uses throughout, a reel about repeating one outfit reached 31,400 people and was saved 288 times, while a much smaller reel about a five minute tidy reached 7,340 and was saved 916 times. Those two figures describe two different things. The first says the hook worked. The second says the post was worth keeping. A strong hook on a thin post produces exactly the first pattern, and then you spend a month puzzled that reach went up and nothing else moved.

Three other things sit outside a hook's reach. It cannot make a post relevant to somebody it was shown to by accident. It cannot rescue audio that is genuinely unlistenable, because the sound-off viewers leave for one reason and the sound-on viewers leave faster for another. And it cannot fix a subject nobody wanted, which is the hardest of the three to accept, because a failed hook feels like a craft problem and is very often a topic problem.

Which gives you a diagnostic worth keeping on a sticky note. If reach is low, look at the hook. If reach is fine and watch time collapses, look at what comes immediately after the hook. If reach and watch time are both fine and nobody saves or sends the thing, look hard at whether it was worth anything to the person who sat through it.

What our software does with hooks

This is the section about the product, and it is one section because the five surfaces and the catalogue above are yours to use whether or not you ever open it.

Synclify transcribes your reels with word level timings, isolates the first three seconds, and classifies the opening against a fixed vocabulary. The vocabulary contains the weak openers as well as the strong ones on purpose. A classifier that only knows good hooks can never tell you that yours is a tired one, which is the single thing you most need told.

proof: 22 named hook types across five surfaces, both closed vocabularies, so an opening it cannot place is reported as unplaced rather than guessed (server/intelligence/taxonomy.ts)

The output that earns its keep is the arithmetic across your own posts rather than the label on any single reel: which openings you have leaned on, how often, and how the ones you almost never use did against the ones you use every week. You could keep that count by hand in a spreadsheet, and honestly some people should.

Questions people ask about hooks