Are you in the TikTok cat niche? What 121,000 videos reveal.

Are you in the TikTok cat niche? What 121,000 videos reveal.

Are you in the TikTok cat niche? What 121,000 videos reveal.

Are you in the TikTok cat niche? What 121,000 videos reveal.

Have you ever wondered how TikTok knows how to show you cat videos while a friend watches them? Call of Duty clips and politics, but no cats?

Choose a category to continue:

you chose selected. Each point here represents a single video about selected.

While you’re in the app, TikTok tracks how you interact with videos. Monitor your watch time, videos you like, videos you scroll through, and comments you leave. Its algorithm uses that information (along with information collected from users who watch similar videos) to adjust your feed, filling it with more videos it thinks you might like.

Selected The videos are usually watched by people who also enjoy it. related topics content. If you see a lot of content on selectedyou’ll probably see more videos on those other topics as well.

Other categories, such as television and movies and sports They end up grouped in different areas.

There are many more videos on TikTok. Too many to understand individually.

We made a map showing where TikTok groups common video topics, based on viewing histories provided by 1,100 users to The Washington Post for this project. Videos shown to people with similar tastes end up located closer to each other, and the content those people didn’t see is scattered elsewhere.

To show you videos you might like, TikTok’s algorithm pulls from a vast map like this, choosing content close to what you’ve already seen.

TikTok users have long wondered how the app’s algorithm learns their precise tastes and places them on a particular “side” of TikTok.

Our map, created from a simplified model of TikTok’s “recommendation system” algorithm, illustrates the platform’s power to show you what you might want and also reveals vast oceans of content that you may never see and may not know exist.

Our analysis shows that gender plays an important role in the videos shown. Humorous videos about everyday life appeal to almost everyone and are grouped on the map, although they do not have any keywords in common. Meanwhile, television, music and sports videos appear in multiple places on the map, showing how users’ tastes teach the algorithm to make detailed distinctions between different teams, musical styles and television genres.

We’ve created a way to explore some of these ‘sides’ for yourself, so you can discover which corner of the social media universe you live in and which topics are unlikely to ever appear on your phone screen.

Explore where hashtags appear on TikTok

The more you see some hashtags, the less likely you are to see others.

For example, the more What TikTok users got, the less they got . The more the users saw, the less they got .

The opposite of #kitten is

Tap a hashtag to explore

More similar

more different

TikTok owner ByteDance reveals little about the precise inner workings of its algorithm. But an article published in 2022 gives a clue: the algorithms used in the company resemble recommendation systems that create a multidimensional map of videos by combining the viewing options of many users. These types of recommendation algorithms have become an industry standard since Netflix first introduced them over a decade ago.

Our version of the algorithm creates a map based on which users watched each video and which didn’t. Unlike TikTok, it doesn’t know hashtags or anything else about the content of the videos, such as text or images. We know our algorithm worked because when we examined the videos’ metadata later, those with the same topic or hashtag tended to cluster together.

TikTok spokesperson Mahsau Cullinane said the Post’s methodology was incomplete and does not “reflect the reality of how our recommendation system works.” He also said the company is “transparent” about how the feed works. The company’s website says that “take[s] takes into account how others interact with the content to help inform their experience” and includes a “non-exhaustive list… of the variety of signals and interactions that the system uses to deliver personalized content.”

Also says Its objective is to “promote[e] a variety of content and topics.” The Post map confirms this: users’ feeds typically come from various places on the map.

For example, here’s a map representing six months of viewing history (about 8,800 videos) for user 462, a woman from Idaho.

Videos that user 462 watched on TikTok

You get a lot of mental health content — about 9 percent of your feed, according to The Post’s analysis.

This group of videos about mental health is interwoven with groups of videos about family and romantic relationships. That means users who watch videos about mental health often also watch videos focused on relationships — TikTok sees the topics as connected. So if people want to get less mental health content, they would also have to skip the relationship-focused videos.

This user also views many other topics. Five percent of their diet is cats and another 9 percent is Taylor Swift content.

Many participants in The Post’s project were apparent fans of the pop star: Dozens, like this user, saw her mentioned in at least 5 percent of the videos in their feed in the summer of 2024, during her Eras Tour.

User 723: a probable Swiftie

The map shows that there is little middle ground in Swift content. If you don’t like the pop star, even if you watch videos about other pop stars or the Kansas City Chiefs, whose tight end Travis Kelce is Swift’s fiancé, you’re likely to get very little content about her. If you get a lot, it’s largely the same content shown on all the other Swifties.

Because Swift’s content is so polarizing, the roughly 1,500 videos about her appear on the map like an island. Other topics with devoted fandoms appear on other edges of the map.

In other research as part of this project, The Post examined how even the most frequent users of TikTok gradually spent more and more time, on average, on the app. Those regular users are more likely to see “storytelling” content, a genre of funny or frustrating stories told in the first person.

This user, who watched about four days of TikTok each month in the summer of 2024, watches many videos set on the storybook peninsula.

User 113: a ‘story’ content viewer

In some stranger examples of storytelling, a robotic voice reads an anecdote, sourced from sites like Reddit, to appeal to the user’s ears, superimposed over unrelated video game images, intended to attract their eyes.

In a way the tiktok algorithm What works is to show users what similar users have liked. That means that some seemingly unrelated topics are located close to each other on the map, because many users enjoyed videos on both topics.

This user, a man who said he is in his 30s and lives near Boston, receives a lot of clips about engineering, but also about the comedy series “The Office.”

User 115

Similarly, fans of #booktok, a home for book discussions on TikTok, often watch content about “Bridgerton” and “Game of Thrones,” both prestige TV shows.

User 168

Just as some unrelated topics end up very close together, some related topics end up very far apart, like music, because not all fans love all types of music.

Other genres of music that this user did not see

This user likes rap music and was shown 271 videos tagged #notlikeus, a reference to Kendrick Lamar and Drake’s beef last year.

But that user didn’t hear much from the pop group or a region about Eurovision, an annual singing competition based in Europe.

The pop music region is near the top of the map, adjacent to the LGBTQ+ videos region, because many users in our sample watched videos on both topics, including those by Chappell Roan, a queer singer with a drag persona.

Your gender has a big influence on your feed.

To the left of the map are videos disproportionately viewed by women, including books and skin care products. Videos on the right are disproportionately viewed by men, including video games and sports cars. Fitness, travel, and work end up in the middle because men and women view them uniformly.

The map algorithm used by The Post was not told anything about the gender of participants, but it discovered on its own that gender is one of the biggest predictors of what’s in your feed. Users who provided their gender to The Post were sharply divided over how their feeds appeared on the map.

Of course, each person’s feed is personalized. There’s nothing stopping you from creating a feed to disrupt the trends of the algorithms, viewing content on one topic and also its opposite, like Dungeons and Dragons and Kim Kardashian, or marriage and “SpongeBob SquarePants.”

About this story

Reporting by Jeremy B. Merrill.

Design and development by Leslie Shapiro and Joe Fox.

Edited by Emily M. Eng, Meghan Hoyer, Yun-Hee Kim, and Akilah Johnson. Copy editing by Courtney Rukan.

Additional support from Caitlin Gilbert, Clara Ence Morse, Alexandra Pannoni, Michelle Jaconi, Emma Grazado, Claire Wallace, Craig Timberg, Dave Jorgenson, Lauren Saks, Carmella Boykin, Joseph Ferguson, Paige Moody, Jake Kara, Scott Derbes, Aaron Brezel, and Katty Huertas.

Noun Project Icons: Comedy by Manda, Family by Fajar Studio, Music by Alina Belogolova, Cat by Indigo Diamond.

Methodology

We use alternating least squares algorithm to make a multidimensional map of videos and reduced it to two dimensions with PacMAP. The map models which videos TikTok offers users; not their commitment decisions. The themes of the map were defined inductively, using keywords. The data, which spans from mid-March to early September 2024, was donated by Post readers.

The “most similar” and “most different” hashtags were calculated from a correlation matrix of the presence of common and non-duplicate hashtags in users’ feeds.

You can read more details about our methodology here.



Reference: Read Latest News in Spanish