This was written as a script for a TikTok video, although I may or may not actually make the video.
Why does TikTok feel so different in 2026 than it did in 2020? I have been thinking about this for a few years, honestly, and I have yet to actually share any thoughts because, well, maybe I'll answer that later...
This is a slightly longer (by longer, like ~2600 words, 5-minute) explanation of the theory, just for the nerds, so if you are not a nerd, feel free to move on.
The reason I am sharing this now is because I sense a growing recognition of just how terrible our media environment has become. "We" (the working-class nerds?) are developing shared, common knowledge showing that these systems not working well for any of us. Many of us have come to agree that they are not serving us in our personal lives, they are not supporting our communities, and they are definitely not serving our larger democratic project. More and more people are seeing this, and are ready to make some change.
And as we pursue change, I think it is helpful (at least for us nerds) to discuss the specific reasons why these systems are failing us. That is the reason why I am sharing this.
Also, a few weeks ago I saw a 10-minute video by Kevin James Thornton, discussing the long, sad story of TikTok's crashout and the decline of social media more generally. And that basically gave me some kind of subconscious permission to share my perspective on it.
Thornton shares some great insights. In particular, he describes his experience of "amassing an audience of people who are interested in a niche topic" and then slowly losing the ability to actually communicate with that audience through the very channel which provided that audience. I have nowhere near the audience size of Thornton and other creators, but I experienced some similar dynamics on a small scale.
Anyway, here are some thoughts.
I will do my best to separate some factual, provable elements of changed from my own hypotheses. I do not have secret access to the TikTok control room (I did request access to the Academic API a few years ago, but they never got back to me), I am just trying to make sense of things.
So, first, some facts.
TikTok got bought. The company literally sold out. More precisely, after years of political pressure, TikTok's U.S. operations were moved to a majority-U.S.-owned venture called "TikTok USDS Joint Venture LLC." That venture seems to be a few billionaires in a trench coat (including Larry Ellison), and ByteDance keeps a 19.9% stake.
The Wikipedia page for selling out includes a musical/artistic definition, "associated with attempts to tailor material to a mainstream or commercial audience." In the context of TikTok, selling out relates more to compromising integrity, morality, authenticity, or principles "in exchange for personal gain, such as money or power."
For TikTok, this was really just a matter of survival. I am not sure what other options they had after years of political pressure: the first threat to ban TikTok came July 31, 2020. A few days later the president signed an executive order which would ban TikTok if it was not sold within 45 days. After a few years of ambiguity, in 2024, Congress passed a divest-or-ban law, ("Protecting Americans from Foreign Adversary Controlled Applications Act"). And In January 2025, the Supreme Court upheld the law. The U.S. venture was finalized in January 2026.
(Side note on this point - I'm intentionally using "the president" and "the administration" because this is one case, among many, where the disagreement is much more top-verus-bottom, elite-versus-plebians, owning-class-versus-working-class, rather than left-versus-right or dnc-versus-gop or republican-versus-democrat. The elite's efforts to buy out TikTok was remarkably bipartisan.)
In a sense, this goes without saying. The algorithms are always changing.
When I was trying my hand at BeInG a CreAtoR in 2020, I developed the following mental model of the TikTok algorithm. It largely aligns with the "batch theory", which is a shared conceptual model of how TikTok's algorithms work at a high level. It goes something like this:
This is a generalization, but the basic idea is that there are gradual, recursive phases of the "instant hit." That is, if your video did well enough in the first 100 random feeds, it will put it in another random pool of, say, 1,000 feeds. There is another competition there among videos that "made it out" of the first round. Then the next round might show it to 10,000 feeds...
It's like a bracket, where each round gives the video more views.
TikTok has shared some details about its systems, which mostly align with the batch theory.
The company blog posts mention specific signals such as user interactions, video metadata, and account settings, which are used to place the videos and evaluate performance.
But the knobs are always in flux -- what Cory Doctorow calls "twiddling." How much weight does watch time get? How much is a comment worth? Does shopping content make a difference, and if so, how much? These are the kinds of features that are always fluctuating as TikTok continues experimenting on us.
In September 2025, the BBC reported that Oracle was "auditing and inspecting the source code and recommendation system underpinning the app, and rebuilding it for US users using only US user data."
So, in addition to the ongoing changes, there may also be some distinct step changes post-sale. It's tricker to determine whether those changes to the algorithm really explain why the app feels worse, or if there is something else. One more fact before I offer my hypotheses...
TikTok was associated with a specific moment in U.S. culture: early pandemic, lockdowns, indoor boredom, the 2020 elections, and the sense that everyone was online basically all the time.
There was something special about TikTok at that time. Whatever made it "special" is somewhat ineffable, but I do think Eugene Wei does a good job breaking down the details. In particular, TikTok and the Sorting Hat describes TikTok as an automatically sorted network based on people's interests, which placed us into invisible subcultures without requiring us to follow anyone. It just analyzed what we watched and liked, and put us in a bucket with similar people.
But our lives and our participation on the app evolved. I enjoyed The Noisy Room website (for the most part), and how it frames this evolution: a few voices became amplified in a way that really distorts our sense of what is happening and what people are actually saying.
Again, early-pandemic TikTok felt like a fun little secret passageway to connect with other people forced online. Now that passageway goes to a loud casino-like room where it seems most people are yelling at us and selling stuff (right now TikTok really thinks I'm going to buy an Apple CarPlay add-on device).
My general hypotheses is that TikTok got bullied into reducing its more subversive features. TikTok created an amazingly effective algorithmic system for creating and discovering social content. That's how it took off: the app provided a fun, functional, collaborative communication system.
Yet after the the political threats and bullying, TikTok was forced to prove it would/could obey the powers that be. It had to show that it could be controlled, that its mysterious recommendation system would not show videos of Momo asking children to attack people (a total conspiracy, by the way).
I propose three prongs of this argument - three changes which I hypothesize TikTok made to their systems: (1) more "manual overrides," (2) decreased randomness/quirkiness, and (3) increased exploitation. Again, these are just hypotheses with supporting evidence, and I can't yet prove them. At the same time, perhaps more importantly, I have not been able to disprove them (but maybe you can help, dear reader?).
There was a sense in the early days of TikTok that anyone could go viral. It was not necessarily unique to 2020 TikTok, but it was palpable. There was a sense of excitement and possibility.
But the record now clearly shows that TikTok is not a merit-based popularity machine. For example, The Markup reported in 2021 about how Tinuade Oyelowo and other creators constantly struggle to figure out TikTok’s changing algorithm. Then in 2023, Forbes reported on TikTok's internal heating button, which they could use to manually boost selected creators/videos.
TikTok has made some noteworthy efforts in response to these critiques. I was especially excited about the transparency center, which was announced in late 2020 and promised to share more information about how the recommendation process works. An important opportunity here was for TikTok to be able to address various folk theories about their algorithm and say bluntly, "no that's not how it works."
But at this point the transparency center, as well as the academic API and related efforts, seems to have been mostly theater. That is, TikTok was just taking performative measures in attempt to prove their app was safe, predictable, and controllable.
One of the knobs that can be "twiddled" in an app like TikTok is the randomness of the recommendations. In technical terminology, this is part of the explore/exploit tradeoff.
Early TikTok seemed to achieve an impressive balance of exploration and exploitation. TikTok's app description, as quoted by Vice, described its ability to "offer the most relevant, interesting, fun, quirky, head-turning videos that you'll never want to stop watching."
"Quirky" is key. These unexpected, off-beat, unusual, eccentric videos laced our feeds in precise, effective doses.
Even TikTok acknowledges the importance of the balance. In Casey Newton's piece about the transparency center, TikTok executives discussed the risk of "repetitive experiences" and the need to interspersed diverse types of videos. They even talked about avoiding filter bubbles! But the point was: random discovery and quirkiness in the feed was really crucial.
This likely requires a bit more of a hands-off approach to content moderation. It entrusts the people scrolling to engage in ways that elevate "quality" content and leave behind inappropriate / undesirable content.
To put it another way, it's easier for a company to review and moderate videos when it's just ten videos that are all getting a million views. But when those ten million views are dispersed across a thousand videos, the company either needs to pay for a lot more content moderation, or, trust their users are ready to separate the chaff from the wheat with their winnowing forks.
My hypothesis -- again, it's just a hypothesis -- is that part of the "retraining" and "rebuilding" of the TikTok algorithm has been a lower dose of exploratory recommendations in the feed. They made TikTok less quirky. The feed still shows plenty of "good," safe content, but the fun surprises are mostly gone.
The flipside of lowering the quirkiness and randomness is a higher dose of "hits" in the feed. Sort of like a radio station playing safe songs over and over again. They are good enough to keep the radio on, but not good enough to make you feel alive.
Again, it's known as the explore/exploit balance. It's a classic dilemma: are you going to go somewhere new, or go back to your same spot? In the TikTok library, the algorithms are balancing similar questions: do we recommend our greatest hits (exploit), or recommend new stuff that we're not sure about (explore)?
Exploration helps a system learn and adapt, while exploitation allows it to "cash in" on what has already been learned. My hypothesis here is that TikTok has leaned into the exploitation side of things.
This "cashing in" often means copying what has worked in the past, whether that be in specific formats, topics, aesthetics, sounds, user interests, etc. The videos and feeds all start to feel more homogenous -- more like slop.
And don't forget about the literal "cashing in" via TikTok shop! Every video you watch, and really, every tap in the app (and elsewhere) can eventually be used to make you buy something. Just keep watching and tapping -- let TikTok handle the rest (don't worry, it will tell you what you want and where you where to click to buy it).
The cycle is somewhat complicated, and it's why this question of "why does TikTok feel so weird now" has no single simple answer. Everything interacts with everything else: recommendation algorithms, creators, audiences, TikTok executives, politicians... It's kind of a mess.
TikTok grew as a fun, quirky, high-volatility video sharing app system that put a spotlight on remix culture and spontaneous creativity during a unique cultural moment in 2020. (Sorry for the run-on sentence). As of 2026, its enshittification cycle appears to have come to a close. After selling out, changing its algorithms, and meeting a changing cultural environment, it lost its magic. The unicorn is alive, but not well.
Notably, it also followed the standard three-step recipe for enshittification. At the beginning (step 1), it was essentially fun and enjoyable for the larger mass audience. Then (step 2), it used that accumulated mass audience to entice advertisers and creators, making their experiences a little better and our experiences a little worse (e.g. with ads and shopping links). Finally (step 3), now in the hands of U.S. owners, the TikTok app is a miserable experience for everyone except TikTok owners.
An important clarification: I am not saying that "TikTok sucks now because the algorithm changed." TikTok does suck now, and its algorithm did change. But it is a larger ecosystem that fostered algorithm changes, ownership changes, and other operation changes that eventually enshittified the app.
I don't know if a fun internet is possible anymore. And maybe fun is not the right word. Perhaps we want more value, or more meaning, or more connection, or more quirk. Personally, I'm not even sure what I want from the internet, if anything, but it's definitely not Instagram or TikTok.
And that's basically the end. I don't even have anything to sell you. I did write a book that is somewhat related, but I don't get commission if you buy it (I get a one-time payment from the publisher that will almost cover all the coffee I bought while writing it). So this post is truly just one nerd trying to share some of what I learned with other nerds out there. I don't even have an alternative app to recommend.
Anyway, I hope you learned something from all this and I would be curious to hear your ideas and questions through an uncorrupted channel.