STORY: TikTok's algorithm has played a key role in the app's major success.

Its unique content recommendation technology is what predicts which videos will pique user's interest.

And it has attracted more discussion than the technology used by rivals, like Instagram and YouTube.

Let's take a closer look at how it works.

Before TikTok, many believed that technology built on users' social connections were the secret sauce to a successful social media app.

Such technology helped Meta's Facebook and Instagram become widely popular.

But TikTok showed that an algorithm that's driven by the understanding of a user's interest could be even more powerful.

Reuters China tech correspondent Josh Ye explains.

"TikTok is a little different from the previous generation of social networks, where algorithms were built on top of what it called a social graph. But then with TikTok, it actually revolves around what it calls interest signals."

While competitors have similar interest-based algorithms, TikTok is able to turbocharge its effectiveness with the short video format.

"If you think about YouTube, the average length of a video is about 10 minutes. So the user interaction with it is a lot more stagnant when it compares to TikTok where you get 10 second videos, right? If you're scrolling through it and you like it or you follow it, that's sending through a lot of different data points to TikTok, which can determine whether this really suits your interest. So the speed at which it collects this data is unprecedented compared to a previous generation or social networks."

But it's not just about feeding users what they want.

TikTok also regularly recommends content that falls outside of users' interest, which the company has said is essential to TikTok's user experience.

TikTok's Chinese owner ByteDance was also an early adopter of the so-called tagging strategy.

"Through our conversations with some of the ex-ByteDance employees, we realized that there's one very unique advantage that TikTok has over the Western social media platforms, which is they focused very early on tagging or labeling every user and as well as every video that's on their platform. And they not only do it through machine vision. But then they also actually hire a bunch of people to manually, laboriously annotate a lot of these kind of content and user profiles on their platform. And as you know, in China the labor cost is a lot cheaper. So ByteDance can afford to be deploying this strategy very quickly and at a large scale or more so than its Western counterparts."