If your Kick numbers moved in May and you did not change anything about your stream, the Kick algorithm change is almost certainly why. On 13 May 2026 Kick started rolling out a personalised homepage that recommends streams to each individual viewer, and it became the default after roughly a week of testing. The front page no longer behaves like a leaderboard of concurrent viewer counts. Discovery is now tied less to raw audience size and more to what the person looking at the screen already watches. Kick also said it would push streams showing artificial or inflated engagement down inside those recommendations. Three separate things happened in that same fortnight, they produce graphs that look almost identical, and the fix for each one is completely different.
Table of Contents
Work out which of the three things moved your numbers
Nobody can help you until you know which one you are looking at. Start here.
One. You stayed still and the channels above you fell. Kick began pushing inorganic activity down the recommendations in the same window as the homepage change. Channels sitting above you on numbers that were never real stopped being shown as often. Nothing about your stream changed at all. Your position in a list did. You see this as a ranking that quietly improved while your own average viewers stayed roughly where they had been for months.
Two. The sort changed underneath you. On the old browse page, a viewer opening a category got the biggest streams at the top and everyone else buried down a very long scroll. Under a recommendation system, that same viewer gets streams matched to their own history instead. If most of your new viewers used to arrive by scrolling a category, that pipe narrowed on a specific date. You see this as new viewers dropping sharply while your regulars carry on turning up as normal.
Three. Some of your traffic was never real. This one is uncomfortable and it is the one worth being honest with yourself about. If you ever bought a promotion, ran a viewer service, or accepted a raid package from someone in your DMs, the same de-prioritisation now applies to your channel. You see this as viewers falling off a cliff while chat activity stays flat, because the number that dropped was never made of people.
| What your graphs show | Most likely cause | What to do about it |
|---|---|---|
| Average viewers flat, but you sit higher in your category than before | Inflated competitors were pushed down | Nothing at all. Keep the same schedule and let it settle |
| New viewers collapsed, returning viewers steady | The homepage stopped sending you browse traffic | Fix your metadata so the recommender has something accurate to match |
| Viewers collapsed but chat messages and follows barely moved | The viewers you lost were not people | Switch off every paid traffic source today |
| Everything slid together across several weeks | Probably not the algorithm at all | Look at your schedule, your game choice and the days you did not go live |
The growth trick that stopped working on 13 May
Every Kick growth guide on the internet tells you the same thing. Pick a quiet category. Something almost nobody streams. Your eight viewers will then sit near the top of that directory page, real people scrolling will see you first, and you snowball from there. It is clean advice and it costs nothing to follow.
You were expecting me to leave it there, weren’t you. Here is where it falls apart. That entire tactic is built on a page that ranks streams from most viewers down to fewest. Once Kick de-prioritised viewer-count sorting on the homepage, the mechanism the trick depends on stopped being the main road a viewer travels. Being first in a dead category only pays when somebody is scrolling that category in the first place. A personalised homepage does not send them there.
Category choice still does something. People browse deliberately when they want one specific game, and picking a category you have no business being in still confuses everything downstream. What died is the idea that category selection on its own counts as a growth strategy.
What the new system can actually see about your stream
Kick has not published the list of signals its recommender uses. Anyone selling you that list is guessing, so do not accept a confident ranking from me either. What you can safely do is reason from what any personalised recommendation system needs in order to function, because that part was never secret.
A system matching streams to individual viewers needs two things from you. It has to know what your stream is. It has to see evidence that people shown your stream stayed on it. Everything inside your control sits under one of those two headings.
- Your category has to be the game you are genuinely playing, updated the moment you switch, because a mislabelled stream gets matched to the wrong viewers and they leave within seconds.
- Your title should contain the words a person who likes this content would actually use, written plainly rather than packed with emoji and abbreviations.
- Your tags describe the flavour of the room. Somebody who watches horror games at two in the morning is a different viewer from somebody who watches speedruns before work.
- Retention is now the honest measure of you. If a hundred people are shown your stream and four are still there a minute later, the system learns something about your channel that no amount of title editing will undo.
- Consistency gives the system repeat evidence to work from. A channel that goes live at the same hours produces a much cleaner pattern than one appearing at random.
Kick said the point of the change was giving smaller and emerging channels more meaningful visibility. That cuts in both directions. Plenty of small channels went up in the same week that other small channels went down, and being small stopped functioning as either an advantage or a penalty by itself.
If you paid for viewers, this is the part to read twice
I am going to stop being funny for a moment, because this bit involves your money.
Inflated viewer numbers on Kick are no longer merely a waste of cash. They are now a signal the platform uses to push your channel down inside the system that decides who ever sees you. The people selling stream viewers have every commercial reason not to mention that to you.
If you have a paid viewer source running right now, switch it off before you do anything else on this list. Then leave the channel alone for a few weeks. Your numbers will look worse for a while, because the fake portion vanishes immediately and the real portion has to rebuild from a smaller base. That stretch is genuinely miserable and there is no way around it. A channel being judged on retention cannot recover while it is being fed viewers who never watch anything.
There is a second reason to care and it is a straightforwardly financial one. Sponsors look at the gap between your viewer count and your chat activity, and so do the people running platform creator programmes. A channel showing a thousand viewers and eleven chatters is a channel nobody pays. The inflated number costs you the exact deal it was bought to win.
What to do over the next four weeks
Kick is a bigger place than it was in April, which changes the arithmetic on all of this. July 2026 was a record month for the platform, with 331,900 active channels and around 593.4 million hours watched. More channels are now competing for the same recommendation slots.
- Write down your average viewers, your chat messages per hour and your new followers per stream for the fortnight before 13 May and the fortnight after. That single comparison usually identifies which of the three causes you are living in.
- Tighten your category and title discipline for two weeks before changing anything else. It is the cheapest change available and it feeds the recommender directly.
- Drop category selection as your growth plan. Stream the game you are genuinely good at talking over, because retention is what gets measured now.
- Watch the first two minutes of your own VOD as though it had been recommended to you by accident. That is precisely the test the system runs on every viewer it sends you.
- Give it a full month before you judge any of it. Recommendation systems work on repeat evidence and one week of data will tell you nothing useful.
Here is the short version of all of it. Your audience very probably did not leave you. The road they used to arrive on got rebuilt in the middle of May, and the map that every growth guide handed you was drawn well before the rebuild started.