What actually makes something go viral, and can it be engineered?

Virality is a property of the spread, not of the content — and while some characteristics reliably increase the odds, the outcome is dominated by chance to a degree that makes reliable engineering impossible.

What the research finds about content that spreads:

High-arousal emotion. Content producing awe, anger, amusement or anxiety is shared more than content producing sadness or contentment. Arousal matters more than whether the emotion is positive or negative — calm and low-energy content spreads poorly regardless of how much people like it.

Social currency. People share things that make them look knowledgeable, funny or virtuous. Sharing is an act of self-presentation, which is the most useful single insight here.

Practical value, since useful things get sent to specific people.

Simplicity and immediacy, because attention is granted in fractions of a second.

A participation hook — something to imitate, argue with, or add to. Formats that invite response outperform formats that conclude.

Why it cannot be reliably engineered:

Extreme outcome distribution. Sharing outcomes follow a power law: almost everything gets almost nothing, and a tiny number get enormous reach. Predicting which item lands in the tail is close to predicting a lottery.

Seeding matters enormously. Studies comparing content spread find that who posts it and when explains much of the variance — the same content from a different account at a different hour performs completely differently.

Cascade dynamics are path-dependent. Early engagement determines algorithmic distribution, which determines further engagement — so small random differences at the start compound into vast differences at the end.

Experiments show low predictability: repeating the same cultural market with different participants produces different winners each time, which strongly suggests inherent quality explains less than we assume.

What is actually controllable: publishing consistently so you take many attempts; optimising the first seconds; posting where an existing audience already is; making the format easy to remix; and understanding that volume plus a decent hit rate beats trying to design one hit.

Manufactured virality exists — coordinated seeding, paid amplification, engagement pods — and platforms actively work to detect it.