How to build a hook for a podcast clip
The first second and a half decides whether a clip gets watched. Here are the hook structures that survive a cold feed, and the four that reliably fail.
Short answer
A hook has to do three things in under two seconds: signal what the clip is about, create a reason to keep watching, and give the eye something to hold. The most reliable structures are the contradiction, the specific number, the mid-sentence entry, and the named stake. Hooks fail when they explain instead of provoke, or when they front-load context the viewer has not agreed to care about yet.
Every clip gets a fair test of about one and a half seconds. The hook is the only part that gets tested for certain.
This is not a copywriting exercise bolted onto the front. It is an editing decision about where the clip starts.
What a hook has to do
Three jobs, simultaneously, in under two seconds.
Signal the subject. The viewer needs to know roughly what this is about, fast enough to decide whether it is for them.
Create an open loop. Something unresolved. A claim needing support, a story with no ending yet, a number without an explanation.
Hold the eye. Motion, a face mid-expression, text appearing. A static frame of someone about to speak is dead air at the exact moment you cannot afford it.
Most weak hooks do the first and skip the other two.
Structures that work
The contradiction. State something that runs against the obvious.
"We spent four months on the launch and the thing that worked was a post I wrote in ten minutes."
The tension is immediate and the viewer stays to find out how both can be true.
The specific number. Precision implies a real story.
"It cost us eleven thousand pounds to learn that."
Eleven thousand works. "A lot of money" does not. Specificity is the signal that something happened.
The mid-sentence entry. Start the clip already in motion, part-way through a thought.
"...and that is when I realised nobody in the room had read it."
Arriving mid-flow creates the sense of having walked in on something, which is a strong reason to stay.
The named stake. Say what is at risk.
"If you get this wrong you lose the account, and you usually do not find out for two months."
The direct address. Works when the audience is narrow and the clip names them.
"If you run a team of under ten people, this is the thing that will break first."
Broad address fails. Narrow address works, because the people it excludes were not going to convert anyway.
Structures that fail
The preamble. "So a lot of people ask me about..." Nothing has happened yet and the viewer has left.
The greeting. "Hey guys, in this clip we're going to..." Every second here is spent on nothing.
The credential. "As someone who has been doing this for fifteen years..." Credentials matter once someone is listening. In the first second they read as throat-clearing.
The generic question. "Do you struggle with productivity?" The viewer has been asked this four hundred times.
The on-screen text layer
Text is read faster than speech is heard, and it works with the sound off. Treat it as a second hook running in parallel.
Rules that hold up:
Short. Under about eight words. It has to be readable in a glance.
Different from the spoken line. If the text repeats the audio word for word, you have used two channels to deliver one message. Let the text state the tension and the voice deliver the substance.
High on the frame. Platform interfaces cover the bottom third with captions, buttons and usernames.
Present from frame one (see also caption styling). Text that fades in at 0.4 seconds arrives after the decision has been made.
Honest. Text promising something the clip does not deliver produces a watch and a swipe away, and the swipe is the signal that matters.
Testing hooks properly
You cannot judge a hook in the timeline. You are attached to the edit and you know what is coming.
Two checks that work:
The cold watch. Leave it a day, then watch it in a real feed, scrolling, on a phone. Almost every weak hook becomes obvious immediately.
The two-second test. Play the first two seconds and stop. Ask someone what they think the clip is about and whether they want to see the rest. If they cannot answer the first, the hook has no subject. If they say no to the second, there is no loop.
The uncomfortable part
A strong hook on a weak clip is worse than no hook. It buys attention and then wastes it, and the viewer learns that your content overpromises.
The hook is not a fix for a moment that was not worth clipping. It is how you give a moment that was worth clipping a fair chance of being seen.
Frequently asked questions
Should the hook be spoken, on-screen text, or both?
Both, doing different jobs. On-screen text can be read faster than speech and works muted. The spoken line carries tone and credibility. When they say the same thing word for word it feels redundant, so the text usually states the tension and the speech delivers it.
Can you add a hook that was not in the recording?
You can move a line from later in the answer to the front, which is normal and legitimate. What you should not do is add words the person did not say or reorder statements so the meaning changes. Rearranging for pacing is editing. Rearranging to alter meaning is misrepresentation.
How long is a hook?
One and a half to two seconds of content. In practice that is one short sentence or a fragment. If it takes a full breath to deliver, it is a preamble rather than a hook.
Do question hooks work?
Sometimes, and they are overused. A question only works if the viewer already wants the answer. 'Want to grow on social?' fails because everyone has heard it. 'Why did the second launch fail when the first one worked?' works because it implies a specific story.
Want your podcast turned into clips that hold?
We cut, caption and distribute short-form for podcasts and founders. Bring one episode and we will show you what comes out of it.
Book a call