Captioning
A/B hook testing
Generating multiple opener variants for the same short-form clip, publishing them as separate posts, and feeding the winning pattern back into the next batch of generations.
A/B hook testing is the practice of generating multiple first-line / first-3-second opener variants for the same source clip, publishing each as a separate post (or to different audience cohorts), and tracking which earns the most retention. The output is two-fold: better immediate posts (the winning hook ships further), and a learned pattern for the next batch (the system biases toward the winning structure). Platforms like TikTok and LinkedIn weight the first-line / first-3-second engagement so heavily that a 20% improvement in hook performance often outweighs a 40% improvement anywhere else in the clip.
Related terms
Hook (short-form video)
The opening 1–3 seconds (audio + on-screen text) of a short-form video, which determines whether feed viewers swipe past…
Retention (video)
The percentage of viewers who keep watching at each point in a video. Platform algorithms use the retention curve — espe…
Brand Voice
An AI feature that learns a creator's actual writing style — vocabulary, hook patterns, sentence rhythm, typical CTA — f…