How many people watch GTA at once? Between 69,568 and 560,212

On 31 July 2026 there were 69,568 people watching live Grand Theft Auto at 09:45 UTC. Twelve and a half hours later, at 22:10 UTC, there were 560,212. Same day, same game, no announcement in between — just the planet turning. An eight-fold swing is the single most important fact about GTA viewership numbers, and almost nobody who quotes one accounts for it.

We sample every live GTA stream on Kick, Twitch and YouTube once a minute. Here is the last 24 hours, five minutes to a point:

The shape is not noise. It is a clean diurnal cycle: a trough between 08:00 and 10:00 UTC, a long climb through the European and South American afternoon, and a peak between 22:00 and 02:00 UTC. Both complete days in our history so far show it — 30 July ran 42,418 to 437,101, and 31 July ran 80,963 to 530,505 on hourly averages. That is a ratio of 10.3x and 6.6x respectively.

I want to be straight about the limits of that claim: our collector has been running for 61 hours. Two days is enough to see a daily rhythm and nowhere near enough to call a weekly one, and the peak on 31 July was itself unusual for reasons I will come to. Treat the shape as established and the exact amplitude as provisional.

Why this matters right now

Take-Two reports Q1 FY27 results on 7 August 2026, the first quarter to include the GTA 6 pre-order window that opened on 25 June, and speculation about a third trailer has been running hot since Rockstar’s website picked up backend changes at the end of July. None of that is confirmed. What is predictable is that the coverage around it will be full of sentences of the form “GTA currently draws X viewers” — and depending on what hour the writer happened to look, X can honestly be 70,000 or 560,000.

Which language has the most GTA viewers?

The audience does not merely grow and shrink. It changes who it is.

We have two independently captured snapshots roughly ten hours apart: a digest built from the hourly R2 bundle at 17:48 UTC on 31 July, and the live endpoint at 04:04 UTC on 1 August.

Language	17:48 UTC	Share	04:04 UTC	Share
Arabic	49,859	16.9%	162,698	49.6%
English	60,978	20.7%	64,794	19.8%
Spanish	4,011	1.4%	46,338	14.1%
Portuguese	10,459	3.5%	41,429	12.6%
Thai	54,478	18.5%	1,700	0.5%
Russian	15,511	5.3%	421	0.1%
German	14,752	5.0%	1,702	0.5%
All languages	294,990		327,699

Thai was the second-largest GTA language on the platform at 17:48 UTC and had fallen to 0.5% ten hours later. Indonesian was 11.0% and Japanese 6.2% in the first snapshot; by the second, neither appears in the top twelve at all, which puts both under 335 viewers. Arabic tripled.

This is the part I find genuinely interesting, because it means the totals are not measuring one audience getting bigger. They are measuring different populations handing off to each other, and the handoffs are large enough to invert the ranking twice a day. We published a post here yesterday reporting that Thai outdrew English. That was true at the moment we measured it, and it is comprehensively untrue now. The correction is the finding.

Which platform has the most GTA viewers?

Same problem, bigger consequences, because platform share is the number people actually argue about.

Platform	17:48 UTC	Share	04:04 UTC	Share	Channels now
Kick	158,341	53.7%	266,618	81.4%	166
Twitch	75,958	25.7%	47,523	14.5%	2,548
YouTube	60,691	20.6%	13,558	4.1%	122

Across the full 24 hours, Kick’s share of all live GTA viewers ranged from 44.1% at 11:10 UTC to 82.8% at 03:45 UTC. Anyone can pick an hour and publish a defensible-looking number anywhere in that 38-point band. Our own post yesterday put Kick at 55.4%, taken when the board held 255,321 viewers — a mid-afternoon UTC reading, and correct there.

Note what the channel counts do to the averages. At 04:04 UTC, Twitch had 2,548 live GTA channels sharing 47,523 viewers — about 18 each. Kick had 166 channels sharing 266,618, about 1,606 each. These are not the same market with different logos on it.

What is actually driving the 04:00 peak

Arabic at 49.6% of the entire global GTA audience comes from 35 channels. That is 1.2% of the 2,836 channels live at the time, and it works out to 4,648 viewers per Arabic channel against 38 per English one.

Nearly all of it sits on a handful of Kick broadcasts, and the stream titles say what is going on. The largest, sxb, was holding 86,184 concurrent viewers — 26.3% of every GTA viewer on Earth at that moment, on one channel — under a title announcing a 24-hour broadcast and an iftitah, an opening. The fourth-largest stream overall, 3mr at 17,171, was streaming the opening of a roleplay city called Vision RP. Arabic-speaking players call the game قراند, “Grand”, and a server opening is an event with a start time.

That is the mechanism behind the 22:10 peak too. A scheduled launch pulls a very large, very concentrated audience into a two-hour window. It also means the 560,212 figure should not be read as a typical night — it is a typical night plus a launch. I would expect the trough to be far more stable than the peak, and so far it is: 08:00–10:00 UTC has sat between 42,000 and 120,000 across every day we have.

So what number should you use?

If you need one figure, use a daily average or a stated-hour figure, and say which. If you need to compare platforms or languages, compare them at the same timestamp — a Kick-versus-Twitch claim built from two different hours is not measuring platforms at all.

Our own numbers carry the same obligation, so: everything above is timestamped, and the two snapshots are ten hours and sixteen minutes apart. I cross-checked the headline language figure against a second slice before publishing — the aggregate says 162,698 Arabic viewers, and summing the Arabic rows in the top-400 stream table gives 162,674, a gap of 24 viewers from small channels that fall below the top 400. Kick’s aggregate and its top-400 sum differ by 56. That is the agreement I would expect, and if it had not agreed this post would say something different.

You can watch the cycle yourself on the live board, see who is holding the audience at any moment on the leaderboard, break it down by language, or read exactly how we sample on the methodology page.

The next real test is 7 August. When the earnings numbers land and everyone reaches for a viewership figure to go with them, check what hour it came from.