When the For You Page Isn't For You: Why Search Beats the TikTok Algorithm for Spotting Trends

There's a quiet assumption underneath how most creators do "research." They open TikTok, scroll the For You Page for twenty minutes, and treat what they see as a read on what's trending. If a sound or a format keeps showing up, it must be everywhere, right?

Not exactly. The For You Page is not a window onto the platform. It's a window onto you — specifically, onto what an inference model predicts will keep you watching the longest. That's a very different thing from what's actually spreading across TikTok, and a brand-new piece of academic research makes just how different it is uncomfortably clear.

In "When 'For You' Isn't For You: Measuring User Agency in TikTok's Algorithmic Feed", researchers Levi Kaplan, Devin Patel, Nicole Gerzon, Alan Mislove, and Piotr Sapiezynski built controlled experiments to measure how much control users really have over the FYP. Their finding, published at the AAAI Conference on Web and Social Media (ICWSM), is that the answer is "less than you think." This post unpacks what that means for anyone using their feed as a trend-spotting tool — and why Kurrently is built on global search instead of a personalized feed.

Is the TikTok For You Page actually personalized for you?

The For You Page is personalized to you, but it's optimized for the platform. The FYP is a recommender system whose objective is to maximize how long you stay watching, and it infers what to show you from implicit signals — chiefly how long you watch each video — rather than from anything you explicitly chose. As the Kaplan et al. paper notes, unlike services where you start by following people you pick, TikTok "begins making inferences based on implicit signals, such as how long they watch particular videos."

The catch is that "what keeps you watching" and "what you actually want" are not the same set. A video you watched twice because it annoyed you reads as interest. A topic you paused on out of morbid curiosity reads as interest. The feed is a mirror of your behavior, not your intentions — and behavior is a noisy, easily-misread signal.

How Kurrently helps: Kurrently doesn't try to guess what you want from how you scroll. You tell it what you're researching with a search query, and it returns what's happening on the platform around that topic. There's no inference layer between your intent and the results, so what you see reflects your actual question, not a model's guess about your watch time.

Why can't you control what the For You Page shows you?

You can influence the For You Page, but you can't reliably steer it, and that gap is the core finding of the 2026 research. Kaplan and colleagues found the FYP algorithm is genuinely sensitive to both explicit signals (like tapping a button) and implicit ones (like watch time) — it does change the mix of content in response. The problem is the durability and discoverability of the controls you have.

Their experiments showed that the single most effective explicit signal — marking a video "Not Interested" — is, in their words, "unintuitively buried in the interface." And even when users found it, the effect didn't last: once accounts stopped actively signaling disinterest, many "find their feeds dominated by such content again." So the one lever that works is hidden, and pulling it once isn't enough. That is the opposite of control.

How Kurrently helps: Control is the default in a search-first tool. When you search a niche in Kurrently, you decide the scope, and the results don't quietly drift back toward something else the moment you look away. You're not fighting an inference model for the right to see a topic — you asked for it, and it's there.

Does marking "Not Interested" on TikTok actually work?

"Not Interested" is the most effective explicit control TikTok offers, but the research shows it's both hard to reach and short-lived. The Kaplan et al. study identified it as the strongest signal a user can send to change their feed — more effective than simply scrolling past content quickly. If you want to actually shift your FYP, that buried button is your best tool.

But "best available" isn't the same as "good." The study found the button is obscured in the interface, so most users never reliably use it, and its effect decays: feeds reverted to unwanted topics after users stopped repeatedly marking them. In practice that means controlling your feed isn't a one-time setting, it's an ongoing chore the interface actively makes harder. For casual use that's a minor annoyance. For research, it's disqualifying — you can't build a reliable trend read on a surface you can't reliably steer.

How Kurrently helps: There's nothing to bury and nothing to repeat in a search-first workflow. You don't have to teach Kurrently to stop showing you something, because it never decided what to show you in the first place. Each search starts clean from your query, so you're never managing the residue of yesterday's scrolling.

Why is the For You Page a bad tool for finding trends?

The For You Page is a bad trend-research tool because it shows you a biased, personalized sample and calls it the platform. Two creators in the same niche will see substantially different feeds, because each FYP is tuned to its individual viewer's watch behavior. Neither feed is a representative cross-section of what's actually climbing on TikTok — they're both slices optimized for retention on one specific person.

That bias compounds in a way that's dangerous for research. If you watched a format last week, the feed shows you more of it this week, which makes it feel like a rising trend when it's really just your own history echoing back. You can't tell platform-wide momentum apart from personalized reinforcement, because the feed deliberately blends the two. The Kaplan et al. work underscores why: the feed is built to amplify inferred interest, not to report ground truth about what's spreading.

How Kurrently helps: Kurrently shows you the platform, not your reflection. Because it searches across TikTok rather than reading your watch history, the trends it surfaces are real platform-wide signals — what's accelerating in a niche regardless of whether you personally have ever watched it. You get the cross-section the FYP can't give you.

Is searching TikTok better than scrolling the For You Page?

For finding trends, searching beats scrolling, because search gives you control over the sample and scrolling hands that control to an algorithm with different goals. When you search, you choose the topic, the niche, and the lens. When you scroll, the feed chooses for you — and as the research shows, it's choosing to maximize your watch time, not to inform you.

The difference is the same as the difference between asking a question and being handed whatever someone thinks will keep you in the room. Scrolling is passive and personalized; you absorb whatever the model serves. Searching is active and intentional; you interrogate the platform directly. For entertainment, passive is fine. For deciding what to spend a day filming, you want the version where you're holding the steering wheel.

How Kurrently helps: Kurrently is built around the active, intentional version. You search the exact niche or topic you create in and read what's genuinely climbing there — not what a feed predicted you'd linger on. The result is research you can trust because you defined its boundaries.

How do you find real TikTok trends without depending on the algorithm?

You find real trends by querying the platform directly and sorting for momentum, rather than waiting for a personalized feed to surface them. The workflow is straightforward: search the niche you care about, look at what's accelerating right now instead of what already peaked, and check adjacent niches for formats about to cross over. None of those steps depend on your watch history, and none of them require fighting a feed for control.

This matters because trends are time-sensitive, and the FYP is a laggy, biased indicator. By the time a format saturates your personalized feed, it has often already peaked for the early movers — and you have no way to know whether it's truly platform-wide or just your own loop. Searching cuts straight to the signal: real videos, real velocity, across real audiences, on a topic you chose.

How Kurrently helps: This is exactly the workflow Kurrently is built for. Search any topic, niche, or sound; see what's trending across the whole platform sorted by what's gaining speed; and explore niches well outside your own without a personalization wall in the way. The same approach that academic researchers had to build custom tooling to achieve, Kurrently gives you by default.

Final thoughts

The For You Page is an extraordinary entertainment engine and a poor research instrument, and the 2026 ICWSM study makes the reason precise: it's built to hold your attention, the controls that would let you steer it are buried, and the content you reject creeps back the moment you stop pushing. A surface you can't reliably control is not a surface you can reliably research.

So stop using your feed as evidence. If you want to know what's actually trending on TikTok, ask the platform directly instead of waiting for an algorithm to decide what's for you. That's the whole idea behind Kurrently — search-first, platform-wide, and yours to steer by default.

Start finding real trends with Kurrently →