Seeing that this topic seems to, rather frustratingly, change every five minutes, let’s look, yet again, at the ever-changing TikTok algorithm (excuse my enthusiasm!). 

In 2020, TikTok’s CEO Kevin Mayer published a manifesto on the importance of transparency for tech companies, especially when it comes to their content algorithms. Mayer committed to being more open than its competitors, indirectly challenging Meta and Google. It’s safe to say that within this topic, TikTok has kept its promise and has some solid documentation on how their algorithm works. 

One topic I’d love to address is TikTok’s strange ‘understanding’ of what the algorithm thinks people want to watch. My partner, a 26 year-old male and other straight male friends in their late 20s and early 30s have begrudgingly downloaded TikTok after being gently pushed by myself or other young women and people in their lives. The moment they joined the app, their feed tends to be flooded with bikini-clad teenagers, crude physical “humour” and what I can just describe as a bunch of British guys acting very lad-y. 

Here, the content TikTok is serving is based on these young men’s demographic data. The algorithm hadn’t had time to work its magic then, but when it did, male friends tended to be more convinced. This is because TikTok collects data on how users interact with different videos. Based on this information, TikTok can determine a user’s interests and serve them related content. TikTok uses the content of each video to understand what topic it pertains to. This is based on the use of hashtags, video descriptions, the TikTok sound used, and the textual spoken audio. 

The platform gets better at tailoring this content for you as you engage with it, but it also bases its recommendations on demographic data such as gender, age, and location.

According to their privacy policy, TikTok adds “inferred information” to your profile, such as age-range, gender, and interests.

Knowing this, it would make sense that TikTok puts audiences into different interest cohorts. By connecting different topics by how closely related they are, TikTok should be able to surface topics you’re likely to enjoy, even if you’ve never engaged with them on the platform before.

For example, if I like crocheting and sustainable home fashion projects, I’m likely into sewing hacks, which means I’m likely into fashion DIY. If I’m into fashion DIY, I’m likely into reversible sewing machine co-ords. Boom, a reversible sewing machine co-ords DIY video reached my feed, and I like and save it. 

TikTok’s transparency policy came about after receiving some criticism around how their algorithm creates echo chambers that promote radicalization and the spread of misinformation. Now some platform representatives have spoken about how the platform is trying to prevent that.

Youtube and Facebook have come under fire for this before, but the truth is that any platform with a content discovery algorithm that relies on engagement is susceptible to creating echo chambers and promoting radicalization. Human psychology tells us that we’re more likely to engage with content that elicits a strong emotional reaction. This incentivizes content creators to promote content that makes us angry or afraid.

TikTok’s answer to the filter bubble effect has been somewhat simple: the platform will show you random content from time to time.

In order to avoid homogeneity of content, the app has started showing users content that they don’t usually engage with. This includes surfacing random hashtags, video aesthetics, sounds, and topics. The app tries to keep things fresh by avoiding content repetition, so you’re unlikely to see two videos by the same creator or using the same sound in a row.

Another interesting incorporation into the algorithm is showing you fresh content that has not had any engagement yet. If you’re a TikTok user, I’m sure you have noticed this.

However, when it comes to how SEO works on TikTok there seems to be a bit of confusion as to where it comes from; video, hashtags, voiceover, caption? Let’s break it down

  • The video’s visuals. According to their privacy policy, TikTok can “detect and collect characteristics and features about the video and audio recordings” by identifying objects, scenery, and what body parts are present in your video. This is used for content moderation and to power their recommendations algorithm.
  • The audio. The platform can process the “text of words spoken” within your videos to further understand what they’re about.
  • Text over the video. Using text over the video also contributes to that understanding of the content. Adding the text natively within the platform might provide a stronger signal, based on the way other content ranking algorithms work.
  • Title and hashtags. This is the OG signal for TikTok and it’s the one they’ve publicly discussed the most. The title and hashtags used in the video help tell TikTok what the video is about, but they can also influence rankings indirectly by affecting engagement and discovery.
  • TikTok sounds. The sound being used in a video is a ranking factor on its own, as it helps the platform understand a video’s content. But the biggest way in which sounds affect your content’s performance is jumping on a trend. Trending sounds get a ranking boost for a short while, since they can predict user engagement.

Locality also factors into engagement. There is one line on TikTok’s official documentation that really caught my eye:

“A strong indicator of interest, such as whether a user finishes watching a longer video from beginning to end, would receive greater weight than a weak indicator, such as whether the video’s viewer and creator are both in the same country.”

There isn’t a lot of clarity about how location is used as a ranking factor, but we know it exists. We can understand that proximity between viewers and creators helps in ranking, but we don’t know at what level this is measured. TikTok tracks user location through SIM card information, IP address, and, if you give your permission, GPS.

Categorized in: