What makes a podcast clip go viral: the eight hook types we look for

A 90-minute episode has maybe six to ten moments worth posting. Finding them is one problem. Knowing why they work is a different one, and it's the one that lets you find them faster next time, or judge whether the tool that found them was right.

Every clip we produce is labelled with one of eight hook types. This is the list, with what each one looks like, because the taxonomy is more useful to a podcaster than to a model.

The two-second problem

The viewer decides whether to keep watching before the clip has really started. They see a face, hear the first few words, read the first caption. If nothing in that window promises something, they're gone.

So the opening line of a clip is doing a different job from the rest of it. The rest has to deliver. The opening has to make a promise. The eight hook types are eight kinds of promise.

The eight hook types

1. Question. The clip opens on a question the listener immediately wants answered. "So what actually happened with the money?" The promise is the answer. Works when the answer arrives inside the clip; fails when it's a rhetorical question the speaker never returns to.

2. Bold claim. A flat statement that's hard to agree with instantly. "Most people should never buy a house." The promise is the argument. The clip needs the reason, not just the line; a claim with no reason is a tweet, not a clip.

3. Absurd. Something that sounds like it can't be true. "I got fired for being too early." The promise is the explanation of how that could possibly be. This is the most reliable hook type for comedy and the most common one for clips that get shared rather than just watched.

4. Stakes. The opening establishes that something matters. "This was the moment I nearly lost the company." The promise is finding out what happened. Stories almost always open this way if they're going to work.

5. Contradiction. Two things that can't both be true, or a speaker contradicting the obvious. "Everyone says do what you love. I think that's the worst advice there is." Related to the bold claim but the energy comes from the tension with something the viewer already believes.

6. Challenge. One person pushing on another. "That's not what you said last time." The promise is the response, and the appeal is watching someone have to defend themselves. Interviews with any friction produce these constantly; friendly chats don't.

7. Story open. "So I'm standing in the airport at 3am." No claim, no question, just a scene that's obviously going somewhere. This one is under-picked by tools because there's nothing quotable in the first line; the promise is entirely in the tone.

8. Reaction tease. The clip opens on someone reacting: a laugh, a groan, a "no way", before you know what to. The promise is finding out what caused it. Reaction content lives here. It only works when the cause is in the clip, which means the clip has to be long enough to loop back.

The rule that beats all eight

Never sacrifice the setup to start on the punchiest line.

This is the mistake every automated tool makes, and it's the mistake we measured ourselves making. Given a joke that started at 49 seconds, our earlier version clipped from 56, because that's where the line with the laugh under it was. Punchiest possible opening, and the clip made no sense.

A hook is the first line of a complete thing. If the punchiest line is in the middle of the bit, the hook is whatever the bit actually opens on, and the bit's setup is the hook. Opening on the punchline gives you two seconds of attention and then confusion, which is worse than no attention.

Practically: when you're choosing where a clip starts, find the strong line, then walk backwards to the sentence where that thread began. That sentence is the start. It's usually less punchy than the line that made you notice, and it's the right place to start anyway.

Which types work for which shows

Not all eight appear in every show, and the mix tells you something about the episode.

Show type	Hook types you'll find	Hook types you won't
Interview with an expert	question, bold claim, contradiction	reaction tease
Two friends talking	absurd, story open, challenge	stakes
Solo, educational	bold claim, question, contradiction	challenge, reaction
Comedy or reaction	absurd, reaction tease, story open	bold claim
Business or founder stories	stakes, story open, contradiction	absurd

If your episodes only ever produce one hook type, that's worth noticing. A show that only generates bold claims is a show where nobody pushes back. A show that only generates story opens has no opinions in it. Both can work, but the second half of the table is a list of what you could add.

Reading a tool's picks

If an AI clipping tool labels its picks, use the labels to audit it:

• Every clip is "bold claim". The model is picking quotable lines, not complete moments. Expect mid-thought cuts.
• No "story open" or "reaction tease" ever. The model is only reading the words. Stories and reactions don't show up in text, so it can't see them. Ours reads audio energy alongside the transcript for exactly this reason; a loud reaction with no speech is invisible otherwise.
• The hook reason is generic. "Engaging content with strong appeal" means the model had nothing. A specific reason quotes what was said.
• The scores all sit in the 80s. Not a ranking. We wrote up why.

The part no hook can fix

A hook is a promise. The clip has to keep it. A great opening on a moment that trails off, or gets cut before the payoff, or runs on for 90 seconds past the point, is a clip that people start and don't finish, and completion rate is what the platforms actually measure.

So the hook decides whether anyone watches, and the arc decides whether they finish. How long a clip should be is really a question about the arc, and it's the other half of this.