Captions, hashtags and being findable

Short-form platforms increasingly work like search engines. What you write around the video decides who finds it later.

A growing share of short-form video is discovered by people looking for something, not scrolling past it. That changes the job of the caption: it is not a place for commentary, it is how your video becomes findable.

Write what someone would type

Use the actual words a person would use when searching for this. Not your internal name for the topic, not clever phrasing — the plain, slightly awkward way real people ask.

If your audience says "what to wear to a winter wedding", write that. The more natural a phrase sounds, the more likely it matches something typed.

Front-load the meaning

Captions are truncated in the feed. The first few words are the only ones many people see, and they carry disproportionate weight.

Put the subject at the start. Context, jokes and asides go after.

The spoken words count too

Platforms transcribe audio. Saying the key phrase out loud in the video helps it be understood and matched, as well as serving the viewer.

Likewise, on-screen text is read by these systems. Burned-in text that states the subject clearly does double duty — it helps the sound-off viewer and it describes the video to the platform.

Hashtags are categories, not keywords

Hashtags help place a video rather than rank it. A handful of relevant ones is sufficient; thirty is not better, and a long list of unrelated high-volume tags can actively confuse the categorisation.

Mix broad and narrow. The broad ones say what field this is, the narrow ones say exactly what it is — and the narrow ones are where a smaller, genuinely interested audience is.

Accuracy is the rule

Tagging something unrelated to catch a bigger audience brings in people who leave immediately, and early exits are precisely the signal that stops a video being shown further.

Mismatched discovery is worse than less discovery.

Write for the person, then check for the search

A caption stuffed with phrases reads badly to the human who actually arrives, and the human is who decides whether to watch.

Write it naturally first. Then read it once more and ask whether the words someone would search for are actually in there. Usually it is a one-word change.