Should you stream on Twitch and Kick at the same time?

Only if a fifth of your audience is worth the trouble.

Thirty-six people were pushing one Grand Theft Auto broadcast to two platforms in the same second this morning. For the median one of them, the smaller platform held 21.6% of their combined audience. That is not a second audience. It is a minority share of the one they already had, sitting in a different chat.

Nobody publishes that number, which is odd, given how much advice is built on top of it. Page one for this question is how-to guides and rules explainers almost the whole way down — Dexerto, Restream, Meld Studio, Streams Charts, Streamster — and between them they will tell you the encoder settings, the bitrate and the terms of service, and not one of them says what happens to the audience afterwards. The Meld Studio guide gets closest, with a claim worth testing: “Most mid-tier streamers (1K–50K followers) now broadcast to at least two platforms every session.” It cites nothing.

We can test it in one game category, because we count every live GTA stream on Kick, Twitch and YouTube once a minute.

The habit is rarer than the guides suggest

Across 16 complete censuses between 04:03:01 and 04:19:01 UTC today, a mean of 2,000 channels were live with this game, holding 298,504 concurrent viewers between them. Thirty-six of those channels — 1.8% — were the same handle running on two platforms at once.

Not most. Not close to most. That is not a like-for-like test of the sentence — its band is 1K–50K followers and we count viewers, so we cannot isolate exactly the group it describes — but the gap between 1.8% and “most” is not the kind of gap a definition closes.

Two caveats, immediately, because they cut in opposite directions. We see three platforms, so a creator simulcasting to TikTok, Rumble or X is invisible to us and 1.8% is a floor. And a creator using a different handle on each platform is missed entirely, which pushes the same way.

One big side and one small side

Here is every one of the 36, placed by how much of their audience the smaller platform held.

The dots do not line up against the vertical axis, which is the first useful thing here: how evenly a creator’s audience divides has nothing to do with how big it is. xQc at fifty thousand viewers splits 75/25 in Twitch’s favour. itsellecee at 138 splits 49.5/50.5, the most even reading in the set. kruelkitty at 111 keeps 98% of it on Twitch.

For 26 of the 36, one platform delivered more than twice what the other did. For 18, more than four times. A simulcast does not hand you two rooms of roughly equal size. It hands you a room and an overflow cupboard, and which one is which is a fact about the creator rather than about either platform.

Creator	Bigger side	Smaller side	Smaller’s share	Language	Mode
xQc	Twitch 37,581	Kick 12,582	25.1%	English	Roleplay
buddha	Twitch 14,844	Kick 5,932	28.6%	English	Roleplay
DarkViperAU	YouTube 2,368	Twitch 1,342	36.2%	English	Speedrun
AnthonyZ	Twitch 3,200	Kick 197	5.8%	English	Roleplay
DGthe99	Twitch 1,575	Kick 274	14.8%	English	Roleplay
harry	Kick 667	Twitch 528	44.2%	English	Roleplay
brunablince	Kick 621	Twitch 371	37.4%	Portuguese	Freeroam
MissBombastic	Twitch 216	Kick 36	14.2%	English	Roleplay
fyzicul	Kick 158	Twitch 33	17.1%	English	Roleplay
Cutieno	Twitch 127	Kick 31	19.5%	English	Roleplay
Silbullet	Twitch 139	Kick 14	8.9%	English	Roleplay
harmless	Kick 82	Twitch 70	46.2%	English	Roleplay

Mean concurrent viewers per platform across all 16 censuses, ordered by combined audience. The other 24 creators run from 61 viewers down to 2.

Add every smaller side together — every person watching a GTA simulcast on the creator’s second platform, anywhere in the world, on Kick, Twitch and YouTube — and you get 21,712 viewers. Seven per cent of the category. That is the whole prize.

What changed in four weeks is not the habit. It is who has it

We ran this same test in the same clock window on 13 August and published it. Then, 22 channels were live on more than one platform, they held 2.4% of everyone watching, the best-ranked of them sat at #11 by concurrent viewers and exactly one made the top 50. The sentence in that post was “they are small.”

Today the simulcasters hold 28.2% of everyone watching. The best-ranked is #1. Four of them are in the top ten and eight are in the top fifty.

The habit itself barely moved — 22 channels out of somewhere between 1,576 and 2,312 live that morning, against 36 out of 2,000 now. What changed is that two very large channels switched it on. xQc and buddha alone account for 23.8 of those 28.2 percentage points. Take the two of them out and simulcasting sits at 4.5% of the category — still a fringe practice, and one that has not quite doubled since August.

Anyone quoting “28% of GTA viewing is simulcast” without that sentence attached is quoting two people.

The reason they switched it on expires on 30 September

Twenty-seven of the 36 are roleplay streams, and the roleplay category is in the middle of a Twitch-funded month. Twitch’s own announcement on 3 September set out GTA RP Week: tune into a nopixel V streamer 8–30 September and collect up to GTA$1,500,000, a Burger Shot tracksuit, three emotes and a stack of chat badges — confirmed by Twitch, and available on Twitch only. NoPixel V’s beta opened on the 8th and took 39.2% of all live GTA viewing on day one.

The effect on the split is not subtle. Among the 33 creators running Twitch and Kick together, the median Twitch share is 69.6% today. On 13 August, on 15 such creators in the same window, it was 31.2%. The same test, twenty-nine days apart, has inverted.

That the Drops campaign is the cause is our inference, not a measured fact — we cannot see why a viewer picked a player. But the campaign is Twitch-exclusive, it is aimed at exactly the content 27 of these 36 creators are streaming, and the flip is enormous. Treat the current Twitch tilt as a paid promotion with an end date on it, because that is what the calendar says it is.

A month of one creator, which is the cleanest reading we have

Sixteen minutes is a short window, so here is a long one. Our 30-day creator board, read at 03:56:41 UTC this morning, carries xQc twice — once per platform — with 36.1 hours of airtime on each, matching to the tenth of an hour. That is a month of the same broadcast going to both places.

xQc, 30 days	Twitch	Kick
Hours watched	1,690,193	467,508
Average viewers	46,863	12,956
Airtime, hours	36.1	36.1
Share of combined hours	78.3%	21.7%

Over a month, with the daily curve and the hour of day averaged out, the second platform holds 21.7% of his watch time. Our sixteen minutes this morning put it at 25.1%. Two different instruments, four percentage points apart, and both land on roughly a fifth.

So what should you actually do

Run the arithmetic on a fifth, not on a double.

If the second platform is going to add something like a fifth again on top of your concurrent audience, the question is whether that pays for the upload bandwidth and a chat you are now reading in two places. Permission is not the obstacle it used to be: Twitch’s own TwitchCon post of 20 October 2023 says “you can now simulcast on any live streaming service”, and every 2026 guide on page one still describes it as permitted. Its simulcasting guidelines do attach conditions — reported consistently across the guides above as a quality-parity rule, that Twitch must not get the worse stream — which we could not verify against Twitch’s own help article, because that page renders in JavaScript and serves no text to a reader that does not run it.

Where the arithmetic lands depends on which side is which. For a creator whose second platform pays the better subscription split — the trade the August piece works through — a fifth more people at a higher rate each may well clear the cost. For a creator at forty viewers, eight more people in a second chat nobody is moderating is a real cost against a small benefit.

What the numbers rule out is the reason most people are given. Simulcasting does not double your reach, and it does not build a parallel audience on the second platform. An audience follows a person to wherever they usually are. Only a minority of it re-learns a new habit because you appeared in a second directory.

And do not pick your primary platform off a category page. Kick holds 57.2% of everyone watching GTA in this window against Twitch’s 41.0%, and yet among the creators actually testing both at once, Twitch is the bigger side for 19 of 33. That gap — category share versus what a platform does for your channel — is the whole subject of the August piece, and four weeks of new data have not dented it.

What this does not show

One clock hour, one game, one morning. 04:0x UTC is when GTA’s Arabic and Portuguese Kick channels are in prime time and much of the Twitch-speaking world is asleep; the category swings more than five to one across a day — 110,080 to 585,687 in the 24 hours to this census. Run this at 21:00 UTC and a different crowd is awake. The 30-day xQc table above is the one figure here with no clock in it.

The Drops confound is unresolved and probably unresolvable. We cannot separate “Twitch is where this creator’s audience is” from “Twitch is paying people to watch this content until 30 September.” Both are true right now.

Identity is matched by handle, then checked. Two different people using the same name on two platforms would be a false pair. For 34 of the 36, elapsed session time agreed on both platforms to the tenth of an hour, and 27 carried byte-identical stream titles; the rest differed by an emoji or an abbreviation of the same title. Two channels were not live on both sides for all 16 censuses — dropping them leaves the median at 21.6%, unchanged.

Kick’s public API gives us the category’s top 100 with no pagination, so a simulcaster whose Kick channel has never charted may never have entered our roster. That censoring removes small Kick channels and, if anything, flatters Kick among the pairs that survive.

Concurrent viewers are not people. We cannot tell whether the same person has both players open, and a viewer parked on a marathon counts the same as one paying attention. Nor is any of this causal: nothing here says what these 36 would draw if they turned the second stream off tomorrow. We measured who is there, not what caused it.

We will run this again on 1 October

The Drops stop on 30 September. The same census, the same 04:0x UTC window, the same handle matching, and the median Twitch share among Twitch-and-Kick pairs either falls back toward the 31.2% we measured in August or it does not. If it holds above 50%, this month bought Twitch something durable and we will say so on this page. If it collapses, the honest summary of September was that Twitch rented the roleplay audience for three weeks.

Citing this

Permalink: https://luciastream.com/blog/should-you-stream-on-twitch-and-kick/

Sampling: 16 complete censuses of every live Grand Theft Auto stream on Kick, Twitch and YouTube between 04:03:01 and 04:19:01 UTC on 11 September 2026, taken from data.luciastream.com/live/streams.json plus live/streams-tail.json — the first file is the top 400 channels and the second is the rest, and you need both. Each census reconciled exactly to live/summary.json with no duplicate channels and no degraded platforms. Mean across the window: 298,504 concurrent viewers on 2,000 channels, split Kick 57.2% / Twitch 41.0% / YouTube 1.9%. A creator counts as simulcasting when the same handle appears on two platforms in the same census; per-creator figures are that creator’s mean across all 16. The 30-day figures come from data.luciastream.com/boards/top-creators-30d.json, read at 03:56:41 UTC the same morning — note that the exporter writes the same 30-day aggregate to the 7d, 30d and all paths, so treat every figure on it as thirty days.

The August comparison is our own published census of 13 August 2026, same clock window, in this post.

Method, platform coverage and known limitations are at /methodology/; the live numbers this is drawn from are at /live/ and the standing rankings at /leaderboard/. If you use these figures, keep the sampling window attached to them — a GTA audience number without a timestamp on it is not a number.