What makes a hook work on talking-head videos?

The short version

The hook headline: specific beats clever

We copied the hook headline from 9,871 videos and sorted its wording. Each bar compares a creator's videos whose headline does this with their own videos whose headline doesn't, with the type of video held fixed.

"Holds up" = clear within the same creator after correcting for testing many things at once, with the creator's previous post showing no such pattern. Example headlines are from breakout videos and link to the original posts.

Skip the warm-up: the first sentence works like a second headline

We transcribed the first six seconds of 12,657 videos and sorted how each one opens. Same test: each creator's videos that open this way against their own videos that don't.

Which headline lever works in your niche

The same comparisons inside each niche. Promising something specific raised saves almost everywhere; what raises views depends on the niche.

Green or red = the 95% range excludes zero. "–" = fewer than 30 creators in the niche used both options. Niches clearly differ on: Text at frame 1 → views, Warning framing → saves. Video editing with AI had too few headline posts to test, and Instagram too few creators.

Descriptions: what to write, and what costs you reach

Same test on the description under the post: each creator's posts with it against their own posts without it. Instagram has too few creators who varied these to test on its own; the results are from TikTok, where saves and shares are reported.

Claims we'll re-test next quarter

These results are new, so on November 7, 2026 we wrote them down before collecting the next data. Next quarter we test each one on new videos only and publish every outcome, including the ones that don't hold up. Each claim is re-tested on posts collected at the next quarterly refresh, leaving out every post used here, with the same code. Same creator, same controls (type of video, hook, production, setting, post age, length). A claim replicates when the new estimate points the expected way with p < 0.05. Every claim is reported, whether it replicates or not.

All registered claims, for both studies: preregistered.json (SHA-256 b70ff7763879a253…).

What to do with this

How we did it

• Same videos as the caption styles study: English talking-head TikTok and Instagram videos from creators in 10 niches, posted 7–60 days before collection (90 on Instagram). See that study for how creators and videos were found.
• Headlines: for the 9,871 videos with a hook headline, a model copied the headline from four frames and classified its wording (form, curiosity gap, specific promise, warning framing, audience, tone, and whether it repeats the first spoken line). Word count, numbers, "you", "I" and emojis were counted from the copied text. 5% were labelled twice; the two passes agreed on 92–98% of answers.
• Spoken hook: the first 6 seconds of audio of 12,657 videos were transcribed with Whisper (word timings), and the opening sentence was classified with the headline alongside. Fields that two passes agreed on less than 85% of the time were grouped more coarsely (opening into "straight to the point", "warm-up first" or "no speech"; matching the headline into matches or different) or left out (whether it teases).
• Descriptions: hashtags, length, calls to action, questions, @-tags and emojis were counted from the post description with fixed text rules, no model involved.
• Cause or coincidence: each result compares a creator's videos that do something with their own videos that don't, with the type of video (teaching, list, story, opinion…), hook style, production, setting, post age and length held fixed, across 11,525 random posts from 1,026 creators. The same test on the creator's previous post, which a headline can't affect, must show no such pattern. Breakout odds compare videos that hit 3x the creator's usual views with a random sample of their posts.
• Statistics: "used by" is the share of creators, each counted once and each niche weighted equally. With many features tested, results must pass a Benjamini–Hochberg false discovery correction (10%) to be called "holds up"; "probably real" clears the usual 95% level but not the correction.

Limitations

• This is observational data, not an experiment. Comparing creators with themselves and holding the type of video fixed removes most of the obvious explanations, but a factor we did not measure could still play a part.
• Transcripts cover the first six seconds only and can contain errors; the detailed opening types are less reliable (79% agreement between two passes), so only coarser groups are tested.
• Description results come from TikTok, where saves and shares are reported; Instagram had too few creators who varied these to test on their own.
• Niche-level results rest on 30 to 100 creators each, so their ranges are wide.
• Results describe English-language talking-head videos in these 10 niches in late 2026.

Questions people ask

Use it, cite it

The aggregated results are free to reuse under CC BY 4.0. Journalists, researchers and AI agents can pull the full numbers below; the plain-text summary contains the whole study.

Cite as

Copy citation

Press questions or a custom cut of the data (a niche, a platform, a time window): contact us. We refresh this study each quarter; this page always shows the latest snapshot.