Sitemap

TikTok Virality: What Makes a Video?

8 min readApr 15, 2025

Authors: Vidusha Adira (Project Lead), Stephanie Pham, Olivia Nguyen, Leah Shin

source

Introduction

As TikTok solidifies itself as the dominant social media platform of this time, it is important to take a look at what content is most popular amongst our younger generations, who are most likely to be on this app. This article is an analysis of the TikTok algorithm and the viral video contents to determine what content is most likely to go viral, relying on the metrics of viewership, shares, and likes. As polarization within the country and the world has been on the rise, this article will also include analysis of how controversy and politics plays into the dominant content of the app.

What engagement features attract the most attention?

The epidemic of 2020 that changed many lives played a key role in the growing popularity of chinese app, TikTok, a social media platform in the form of short videos. In relation to its easy user interface, engagement features, and addicting algorithm, it’s no wonder the app has become as popular as it is today. The app allows users to easily interact through video views, likes, shares, downloads, and comments, creating a platform where people can share thoughts, content, and opinions. Let’s take a look at what engagement features determine the virality of a TikTok video.

Correlation heatmap of Engagement Metrics

The heatmap displays the correlation between different engagement features. High correlation in the bright pink squares are between likes and downloads, likes and shares, and downloads and comments. So, these metrics increase together, raising their importance when measuring engagement and popularity of each video. Downloads and likes stand as influential factors when measuring virality, where likes represent viewer appreciation and downloads represent the value of the content being saved.

Videos with high comments tend to be downloaded and saved more often to be sent to friends and talked about. These videos prompt discussion and are more likely to be spread throughout the platform. Another high correlation is between views, an easy way to establish the popularity of a video as users tend to rewatch their favorite content, and likes, a quick and easy way to save videos into a folder on the app. Although they’re strongly correlated as shown in the pink box above, it’s often seen that viral videos get a larger amount of views first, and eventually gain more likes as it resonates with more and more people.

In general, when we talk about the virality of a video and its success amongst the audience, there’s multiple ways to measure virality using engagement metrics. With every viral video, users will be engaging with both the content and each other. It can be seen from the heatmap that likes and downloads are the engagement metrics that have the highest correlation to other metrics, reflecting immediate user interest and long-term attention.

Keywords associated with viral TikToks

Next, to better understand whether certain keywords could trigger virality, we decided to explore transcripts of viral TikToks. The transcripts were taken from the “Tik Tok User Engagement Data” data set on Kaggle, published by user Yakhyojon. To distinguish between viral and non-viral content, we applied a virality threshold, defining a video as “viral” if it accumulated at least 100,000 likes. This threshold allowed us to focus our analysis on content that had demonstrably captured wide attention and engagement.

.

Shown above is a word cloud that provides insight on the most common used words found in the transcripts of viral Tik Toks. Based on the results, we can see that “claim,” “revealed, “discovered,” “learned,” and words associated with knowledge are correlated with viral Tik Tok videos. This association indicates that viral Tik Tok content often revolves around spreading knowledge or discovery. This aligns with the rising popularity of using Tik Tok as an informal search engine and news source and not solely for entertainment purposes.

Additionally, we can see that “friend,” “colleague,” “someone,” “internet,” and “online” are also highly represented in the word cloud . This indicates that these viral videos are published by individual creators who have repackaged information from other sources and platforms, rather than traditional news outlets, such as ABC or CNN, where information undergoes multiple layers of fact-checking and professional journalism processes. Viral TikTok content thrives on personal storytelling and informal “word-of-mouth” credibility, making it feel more relatable and conversational, thus further enhancing engagement.

How does the sentiment of content affect engagement?

To see what type of emotions typically drive forward engagement, we analyzed the sentiments of these videos based on their transcripts and captions to gain deeper insights into if negative content is viewed as more entertaining by those viewing it.

Taking a look at the amount of videos that go viral, we can see that positive videos are far more likely to go viral than negative ones, demonstrating an inclination overall for viewers towards positive sentiments rather than negative. An explanation for this could be based on the engagement metrics, and the different ways the negative content is interacted with in comparison to positive.

Breaking down the videos that do go viral, we can see from the violin plot above that compares views to sentiment, the positive does have a much higher peak, meaning that the highest viewed video is one with a positive sentiment. However, the peak of the views for the negative and positive sentiments are clustered around the same values, with the views median of the positive sentiment only being slightly higher than the negatively inclined videos. This indicates that negative videos that go viral gain the same amount of views as positive ones.

The violin plot for the shares in comparison to sentiment follow similar trends to the views, with the shares peak of the positive sentiment videos being much higher in comparison to the negative videos. However, the peaks of the negative and positive videos are both clustered around the same values, with the median of the negative videos even being slightly higher than the positive. As shares are not an indication of a viewer’s like or dislike for the video they are watching, this slight edge negative videos have over positive ones could be due to the fact that negative videos might elicit stronger reactions, whether that be agreement or not, causing viewers to share to discuss their contents.

Again, the comments versus sentiment plot follows a similar pattern to the ones before. As the median values are close to the same, there is no indication that negative videos might encourage stronger emotions, whether they are agreeing or not to the content, and therefore more likely to engage with the content. To determine if positive content is truly more popular as would be assumed, we then took a look at the likes each video has.

The likes versus sentiment violin plot also followed the same pattern as the previous ones, with the median values almost the exact same. Overall, these plots indicate that while negative content is much less likely to go viral, once it does, it does not perform much differently than positive content. With its similar performance, it can be then inferred that this negative content does not go viral because of a negative sentiment towards it, as might have been assumed as content that elicits a stronger response is assumed to be more likely to be shared and commented on, therefore boosting engagement and pushing the content to more users.

Controversy in TikTok

To explore the impact of controversy in the success of a TikTok, we took a look into how a video’s engagement metrics are affected by whether it is a “claim” or an “opinion” — an “opinion” being an individual’s personal thought while a “claim” is from an unverified/unidentified source. To compare, we scoped three key metrics: view count, share count, and like count.

The views plot displays the vast majority of opinions rack views in the thousands to ten-thousands range, while a majority of claims range in the hundred-thousands to millions.

The likes plot illustrates that most of the opinion TikToks receive around one hundred to one thousand likes, while most of the claims receive around five-hundred thousand to one million likes.

The shares plot reveals that most opinion videos achieve around one hundred to one thousand shares, with most claim videos receiving around ten-thousand to fifty-thousand shares. However, out of the three metrics, video shares show the smallest disparity between claims and opinions.

The above distributions display similar patterns, with claims outperforming opinions by having a greater number of views, likes, and shares. This suggests that non-opinion content, such as reuploads, media coverage, or unverifiable claims, tends to have higher virality. The controversial nature of certain claims, particularly when they involve misinformation or emotionally charged topics, may gain more traction on TikTok’s platform. The results underscore how controversy fuels visibility and interaction, reinforcing the idea that claims — regardless of accuracy — often spread faster and reach larger audiences than personal opinions. This raises important questions about TikTok’s content amplification mechanisms and how virality is shaped by the nature of the content rather than its credibility.

Conclusion

Our analysis of TikTok’s metrics and the interplay of controversy, sentiment, and keywords has demonstrated how this social media app drives engagement and amplifies discussion. Ultimately, the platform drives engagement through controversy and relatability, indicating a reliance on emotionally charged narratives. Understanding how its algorithm works is crucial to understanding how content spreads and influences online discourse, helping creators, policy makers, marketers and more navigate its everyday impact.

--

--

No responses yet