What language do most GTA streamers speak? English, by far

10 min read

languagekickarabicdatagtacreators

English, and it is not close. Across ten complete censuses of every live Grand Theft Auto stream on Kick, Twitch and YouTube between 04:12:36 and 04:22:34 UTC this morning, an average of 1,221 of 2,126 live channels carried an English language tag — 57%. Portuguese was second on 489 channels and Spanish third on 219, though those two swap places the moment you rank by audience instead, which is a small hint at what the rest of this post is about.

That is the honest answer to the question as asked, and it is the least interesting number in this post. Ask a slightly different one — not what language GTA is streamed in, but what language it is watched in, and over thirty days rather than one hour — and the ranking turns inside out. The language that wins is live on 1.5% of the channels.

Who is talking

Language Live channels Share Viewers Viewers per channel
English 1,221 57.4% 198,484 163
Spanish 219 10.3% 64,362 294
Portuguese 489 23.0% 44,113 90
Arabic 31 1.5% 23,739 766
Thai 30 1.4% 6,682 224
French 47 2.2% 1,996 43
Russian 19 0.9% 1,846 96
Everything else 71 3.3% 2,256 32

English is a majority of the channels and a majority of the viewers, so on a snapshot the question looks settled. The column that gives the game away is the last one. An English-language GTA stream had 163 people watching it, on average. An Arabic one had 766.

Who is being watched

One clock hour is a bad way to measure a global audience, and we have proved that on ourselves. On 31 July we published a snapshot in which Thai led on 21.3% of viewers, Arabic was second on 18.4% and English was third on 16.6%. Every one of those figures was true of the minute it was taken, and not one of them survives a different minute. GTA’s audience rotates with the world’s time zones; whoever is awake when you look, wins.

So count hours instead of heads. boards/top-creators-30d.json ranks the hundred most-watched GTA creators on the planet and carries, for each, the hours watched they accumulated over the last thirty days. Grouped by the language they broadcast in, as of 03:57:55 UTC on 12 September 2026:

Language Creators in the top 100 Hours watched, 30 days Share
Arabic 29 21,244,094 49.7%
English 30 7,314,842 17.1%
Polish 6 4,606,570 10.8%
Thai 2 2,804,866 6.6%
Spanish 9 1,767,067 4.1%
French 3 1,542,592 3.6%
Portuguese 5 1,396,305 3.3%
Everything else 16 2,060,828 4.8%

Arabic. Twenty-nine creators, and very nearly half of everything the top hundred earned between them — almost three times English’s share off roughly the same number of creators.

A leaderboard share is a weak claim on its own, because the denominator is the leaderboard. This one does not have to stay weak: we sample the whole category every sixty seconds, so the real denominator is available. Total GTA viewing across Kick, Twitch and YouTube over the thirty days ending 04:00 UTC this morning came to 184,459,273 viewer-hours. Those twenty-nine Arabic channels account for 11.5% of it on their own. The thirty English channels on the same board account for 4.0%.

Eleven and a half per cent is a floor, not an estimate. It counts only the Arabic creators big enough to place in a global top hundred and ignores every Arabic channel below them.

GTA is streamed in English and watched in Arabic Share of live channels (12 Sep 2026, 04:0x UTC) against share of 30-day hours watched among the top 100 creators 0% 10% 20% 30% 40% 50% 60% share of live channels share of 30-day hours watched English 57.4% 17.1% Arabic 1.5% 49.7% Portuguese 23.0% 3.3% Spanish 10.3% 4.1% Polish 0.0% 10.8% Thai 1.4% 6.6% French 2.2% 3.6% Source: luciastream.com. Channel shares from ten complete censuses of every live GTA stream on Kick, Twitch and YouTube; hours watched from boards/top-creators-30d.json, which ranks the top 100 creators by average viewers.

The Polish row is the cleanest demonstration of why the two halves of this post disagree. Six Polish creators hold 10.8% of the top hundred’s hours watched, more than Spanish and Portuguese put together off eight fewer channels. In all ten censuses this morning, the number of Polish GTA channels live was zero. Not small — zero. They were asleep. Written from the 04:00 UTC census alone, this post would have concluded that Poland does not watch Grand Theft Auto.

Every one of them is on Kick

All twenty-nine. Not a majority, not “mostly” — twenty-nine out of twenty-nine, with nothing on Twitch and nothing on YouTube. The English thirty are spread across all three platforms. The live census says the same thing about the rest of the language: 99.97% of this morning’s Arabic GTA viewing was on Kick, and the entire Arabic audience on Twitch averaged seven people.

The top of that list is a channel almost nobody writing about GTA in English has heard of. 3mr accumulated 5,976,066 hours watched in thirty days — more than four times the entire Portuguese-speaking contingent of the same board, and nearly three times what xQc managed as the biggest English-language GTA creator by hours watched in the same window. He did it by being live for 446 hours, roughly fifteen a day, at an average of 13,400 concurrent.

sxb is the opposite shape and the more startling of the two. He was live for 23.4 hours in the entire month and averaged 138,931 viewers while he was, which is how 23 hours becomes 3,244,056 hours watched. His peak is the largest single GTA stream we have ever measured: 215,352 concurrent, at 17:33 UTC on 1 August 2026.

This is not a historical curiosity either. At 04:00 UTC this morning the biggest single GTA stream anywhere in the world so far that day was drb7h at 132,300 concurrent — Arabic, on Kick, and absent from every English-language GTA roundup we can find.

Why the two answers disagree

The two measurements are not in conflict; they are describing differently shaped audiences. English GTA is a long shallow tail — over a thousand channels averaging 163 viewers each, most of them very small. Arabic GTA is a few dozen channels, and at the top it behaves less like streaming than like appointment television: a small number of broadcasts that a very large number of people show up for at once.

Both shapes can produce the same hours watched. Only one of them produces a big number when you count channels, and the English-language press counts channels — or rather, it counts the channels it already reads.

What this does not show

The thirty-day table is a top-end sample, not a population. That board ranks creators by average viewership and stops at a hundred, so “49.7% of hours watched” is a fact about those hundred creators and nothing more. English’s enormous tail of small channels is excluded by construction, and it is larger than anyone else’s tail. Do not read the table as “half of GTA viewing is in Arabic” — the population-level figure is the 11.5% above, and even that is a floor for the opposite reason.

Our view of Kick is truncated, and it cuts the way you would not expect. Kick exposes roughly its top hundred channels to discovery while Twitch pages to about a thousand, so we see few small Kick channels. That understates Arabic channel counts — Arabic is on Kick — and flatters English and Portuguese in the census table above. It does not touch the hours figures, because the channels driving those are far too big to be missed.

Language is a tag, not a transcript. We record the language the broadcaster sets on the channel. A stream that is spoken in Arabic and tagged English is filed as English, and we have no way to catch it.

We deliberately have not told you what these streams are. Our classifier files all twenty-nine Arabic channels as freeroam rather than roleplay, and that is very close to meaningless: freeroam is the classifier’s default bucket, and the classifier reads title keywords, which it does far better in English than in anything else — a census last month put its non-English hit rate at 17% against 39% for English. Somebody should watch these streams and write down what is actually happening on them. We can tell you how many people are there; we cannot honestly tell you what they came for.

Concurrent viewers are not people, and viewer-hours are built from hourly means, so a short sharp spike reads lower than its instantaneous peak. The census covers one clock hour and the board covers thirty days, on purpose — that gap is the whole argument, not a defect.

Why this matters on 19 November

Grand Theft Auto VI ships in sixty-eight days, and essentially every forecast of what its launch will do to streaming has been written from inside the English-Twitch-roleplay frame: which NoPixel names will switch, what the roleplay scene does without a PC version, how Twitch’s directory reshuffles.

That frame is missing most of the watch-time. The largest block of GTA viewing we can pin to named channels is twenty-nine Arabic-speaking ones on a platform that barely features in the coverage, and the biggest single GTA stream we have ever recorded belongs to one of them. On launch night, the largest audience in the world for Grand Theft Auto is at least as likely to be watching in Arabic as in English. It already has been, more than once.

If you want to watch that happen rather than take our word for it, the live board is filterable by language, and the Arabic one is its own leaderboard.

Check us

Every census figure above is the mean of ten complete censuses of every live GTA stream on Kick, Twitch and YouTube between 04:12:36 and 04:22:34 UTC on 12 September 2026, taken from data.luciastream.com. In all ten the top-400 board and the tail file carried matching timestamps, the per-stream counts reconciled exactly to the published totals, no channel appeared twice, and no platform reported degraded coverage. The thirty-day figures come from boards/top-creators-30d.json at 03:57:55 UTC the same morning, and the 184,459,273 viewer-hour denominator from the hourly series in live/summary.json.

Permalink: https://luciastream.com/blog/what-language-do-gta-streamers-speak/. If you use any of this, the sampling window matters as much as the number — quote both. How we collect it is written up in the methodology, and the raw endpoints are public if you would rather check than cite.


Numbers in this post come from our own sampling — seehow we measure.