# How to Find Viral Videos Before They Become Saturated

> Watch a fixed sample of creators and searches in your niche over time, revisit the same posts to measure how fast views are growing, and check whether new, independent creators are still succeeding with the structure. Then ask what useful variation you can add. One big view count or a repetitive feed cannot tell you whether the window is still open.

By Everlyn Weng, Co-founder of AdAnt. Updated 2026-10-20.
Canonical: https://adant.ai/guides/spot-viral-trends-before-saturation

Nobody can reliably predict the next viral video. What you can do is decide well when to join a pattern you can see forming, and when to leave it alone. Sometimes the right move is to bring a familiar structure to an audience nobody has served with it. Sometimes it is to pass.

## What does it mean for a trend to be saturated?

**A trend is saturated for you when similar videos are plentiful compared with your audience's interest, and you have little useful to add.** That depends on the audience: a crowded joke can sit beside a practical problem that the same structure has barely touched.

Three things move at different speeds, and it helps to track them separately:

- **Topic interest:** people care about a problem or an event.
- **Format adoption:** more creators use a similar storytelling sequence.
- **Execution fatigue:** viewers keep seeing interchangeable versions.

A popular topic can support new treatments long after one execution has gone stale, and a demonstration format can stay useful after the meme sound that carried it fades.

## How do you build a sample you can track over time?

**Fix the platform, country, language, audience and problem first, then keep the same creators and searches from one review to the next.** A global trend chart and English-language Reels about organizing a rented kitchen are different research objects.

A practical routine: ten creators and five audience-problem searches, reviewed twice a week. That is manageable, not a representative survey. Save ordinary posts alongside the standouts and log any creator you add or drop; otherwise a change in your sample looks like a change in the market.

Discovery tools can add leads. TikTok's [Creative Center](https://ads.tiktok.com/business/creativecenter/pc/en) has a Trends section, and Creator Search Insights surfaces content gaps: topics "searched for often, but aren't featured in a large number of videos on TikTok" ([TikTok](https://newsroom.tiktok.com/creator-search-insights?lang=en)). A content gap is about topic supply and demand, though; it does not tell you whether a particular video structure is saturated. A trending hashtag or a busy discussion still needs checking with the video audience you want to reach.

## How do you measure whether a video is still growing?

**Look at the same post twice and divide the change in views by the time between the two looks.** Without at least two observations you only know a running total, not how fast it is growing now.

> View velocity = change in visible views ÷ time between the two observations

A video that goes from 20,000 to 32,000 views in 24 hours added about 500 views an hour over that interval. (Example numbers.) That is an average for the interval, not a forecast for the next hour, and rounded public counts make it approximate.

Keep timestamps, the timezone and screenshots. If a count goes down or jumps oddly, check for rounding, an edit or a recording mistake; do not quietly correct it to the story you expected. Record fresh observations of older candidates as well as new posts: an old hit still gaining views does not show that newcomers are succeeding.

## How do you decide whether the window is still open?

**Answer five questions, each with its own evidence, instead of combining them into one score.** Strong growth cannot make up for poor audience fit or a product you cannot show.

| Question | Evidence to look for | What it means for you |
| --- | --- | --- |
| Is interest moving? | Repeated observations of posts and related questions | Investigate patterns with current activity |
| Is the structure spreading independently? | Different creators, original footage, separate executions | Prefer ideas that transfer over repost chains |
| Are new versions still standing out? | Recent posts compared with their own creators' baselines | Check the opportunity reaches beyond the first winners |
| Can we add something useful? | A new demonstration, audience, use case or credible point of view | Avoid an interchangeable copy |
| Can we ship while it is still relevant? | Assets needed, review time, production limits | Pick a feasible entry, or pass |

Write a short evidence note for each answer. Label weak or missing evidence as such, rather than turning it into a zero or an invented estimate.

## What stage is the trend in?

**Sort what your sample shows into four working states, and match the response to the state.**

| State | What your sample suggests | Reasonable response |
| --- | --- | --- |
| Emerging | A few examples with encouraging but limited evidence | A small exploratory test |
| Growing | Several independent, recent versions standing out | Test a distinctive adaptation soon |
| Crowded | Many near-identical versions, mixed recent results | Only with a stronger contribution or use case |
| Declining or unclear | Weak recent evidence, or observations that conflict | Monitor, adapt the underlying structure, or pass |

These are editorial labels, not platform ones. No number of videos, sound uses or days marks saturation. Don't call a decline because a few copies failed (their audiences, quality and timing differ), and don't assume one late success means every brand still has room.

## What does entering early look like in practice?

**Enter with the problem the trend solves, not a copy of the trend.** Here is an example, with every number and observation hypothetical.

On a watchlist of small-apartment creators, the first review turns up two independent cupboard transformations doing better than their creators' usual posts. At the next review there are four more: one stands out and three do not. That calls for investigation, not a verdict. Look at whether the standouts share a clear problem, a visible installation and a measurable result, while the weaker ones only show a nice finished cupboard. Don't assume those differences caused the results; use them to design a test.

A storage brand could add what is missing: an organizer installed in a genuinely shallow rented cupboard, with the dimensions shown and the door closing on camera. The format is familiar, but the proof answers a buyer's real question.

If filming that proof takes three weeks and the concept leans on a short-lived sound joke, the entry may no longer make sense. If it addresses a problem people always have, the demonstration may stay useful after the sound fades.

## What does chasing a trend cost you?

**Measure the time from spotting a reference to approving a usable draft, including reshoots and review.** Speed only helps if the result still makes a credible promise.

Before making a batch, run the smallest test that could change your decision. One original version may expose a production problem; several comparable versions help judge whether the direction deserves more. Neither proves a universal cause.

Keep reach, audience growth and business results separate. Our [49.9M-view Instagram case study](https://adant.ai/guides/instagram-49m-views-one-month-case-study) spells out what its views can and cannot prove, including revenue. Apply the same discipline to your own tests.

## How should a trend review end?

**With a dated decision: test, monitor, adapt or pass.** Name the evidence that would change your mind and when you will look again. That turns trend watching into a creative process you can repeat.

For the formats that are working this month, see [viral content formats](https://adant.ai/guides/viral-content-formats). To research your own niche, [AdAnt's social content strategy](https://adant.ai/ai-social-content-strategy) collects recent videos with their sources and flags outliers; ask it for counterexamples and a feasible original contribution, not just the biggest videos in the category.

## Frequently asked questions

### Is it too late if a trend already has millions of views?

Not necessarily. Check whether recent, independent versions still stand out and whether your audience has a need nobody has answered. Total historical views do not tell you how much opportunity is left.

### How many examples prove a trend is real?

No fixed number proves it. Independent examples count for more than reposts, but a small sample is still uncertain. Use it to justify a bounded test, not to predict a result.

### Do trends always move from TikTok to Instagram?

Do not assume a fixed direction. A pattern can start independently on each platform or reach different communities at different times. Validate each platform with its own examples and your own results.

### Can software guarantee early trend detection?

No. No tool can guarantee a candidate will spread or work for your account. Good software makes the observation window, the sources, the comparison method and the missing data visible, so you can check its reasoning.