How to go viral on TikTok, according to 1,328 videos that did
Most advice about going viral is repeated opinion. These are counted findings from 1,328 viral short videos we analysed frame by frame: 56.3% contain no spoken words at all, 53.2% open on a visual hook rather than a line of dialogue, 69.4% ask the viewer for nothing, and the median viral TikTok is built from just 5 cuts. The pattern is a strong first frame, fast visual storytelling, and no ask at the end.
Most of what goes viral has nobody talking in it
747 of the 1,328 videos in our cohort — 56.3% — produced no speech at all when we transcribed them. On TikTok specifically it is 66.3% (472 of 712).
That is the finding with the biggest practical consequence, because almost every piece of "hook script" advice assumes you are saying something in the first three seconds. In most of what actually spreads, the first three seconds are an image and a piece of on-screen text over a sound.
The first frame does the work, not the first line
53.2% of the videos opened on a visual hook: a frame that raises a question before any word appears. A bold claim came second at 28.3%, and the direct question — the hook most commonly recommended in guides — appeared in only 2.3%.
Practically: decide what your first frame shows before you decide what it says. If the opening frame could be swapped for a stock image without loss, it is not doing anything.
Cut more often than feels comfortable, but not endlessly
The median viral TikTok here has 5 cuts (mean 7.4); Instagram Reels 7; YouTube Shorts 31, because Shorts run much longer. 46.4% were classed as fast-paced — but 32.2% were slow, so a slow, held shot is not automatically a mistake.
Format follows from that: b-roll montage was the most common at 32.5%, ahead of talking head at 22.2%.
Stop asking for the follow
69.4% of these videos — 922 of 1,328 — carry no call to action of any kind. No "follow for more", no "link in bio", no "comment below". Where a CTA did appear, the most common was "comment below" at 17%; "follow for more" was only 6.2%.
We cannot prove the absence of a CTA caused the reach. What we can say is that the single most repeated piece of short-video advice is missing from roughly seven in ten videos that broke out.
You do not need followers first
Across 2,593 captured videos with a verified follower count, the median video was played 45.7 times its creator's follower count, and 76.5% reached at least ten times their following.
Accounts under 10,000 followers did it more reliably, not less: of 737 such videos, 95.9% passed ten times their follower count. Distribution on these platforms is decided per video, which is why a first post can outrun an established account.
What this evidence cannot tell you
These are the shared traits of videos that already spread — not a controlled experiment, and not proof that any one trait caused it. The cohort is 1,328 full-fidelity analyses captured between 20 April 2026 and 9 September 2026, from fifteen subject areas and English-language queries. The full tables, denominators and method are published on our research page.
Read them as a description of the shape of what works, then test the shape on your own subject.