How many watched the GTA 6 Extended Look? 959,078 at once
17 min readUpdated 29 August 2026

Nobody outside Netflix knows how many people watched Grand Theft Auto VI: An Extended Look on Netflix. Netflix does not publish concurrent viewers, and as of this morning neither Rockstar nor Netflix has released a number of any kind.
Here is the number that does exist. At 19:25 UTC on 27 August 2026, twenty-five minutes into the premiere, 959,078 people were watching a live Grand Theft Auto stream across Kick, Twitch and YouTube at the same moment. That is the highest reading in the thirty days we have been sampling, and it is 48% above the previous record we published.
Ninety minutes later it was gone. And when Rockstar put the whole thing on YouTube for free, six hours after that, nothing happened at all.
What this number is, and what it is not
It is not the Extended Look’s audience. We do not measure Netflix, and we cannot see how many people had the video open in a browser tab.
What we measure is every live Grand Theft Auto stream on Kick, Twitch and YouTube, sampled once a minute. During the premiere window, the overwhelming majority of that audience was watching creators react to the Extended Look as it aired — Rockstar had cleared creators to livestream and react to it the moment it hit Netflix, provided they added their own commentary. So this is the co-viewing audience: the people who chose to watch the reveal with someone rather than alone.
That is a real and separately interesting number, and as far as we can find it is the only cross-platform concurrent figure anyone has published for this event. Destructoid’s live blog is reported to have counted a combined 881,000 viewers across Twitch GTA-reveal streams shortly before the premiere — we have not verified that figure ourselves, and it covers one platform at one moment, but it is the closest comparison in circulation.
The shape of it
| Time (UTC) | Total | Kick | Twitch | YouTube |
|---|---|---|---|---|
| 18:00 | 339,484 | 172,117 | 96,238 | 71,129 |
| 18:30 | 388,253 | 209,569 | 77,966 | 100,718 |
| 18:55 | 628,854 | 232,688 | 81,200 | 314,966 |
| 19:25 | 959,078 | 175,662 | 77,019 | 706,397 |
| 19:45 | 770,943 | 217,144 | 72,912 | 480,887 |
| 20:30 | 494,575 | 243,967 | 44,143 | 206,465 |
| 21:00 | 527,130 | 292,095 | 84,441 | 150,594 |
| 22:00 | 458,426 | 295,448 | 91,714 | 71,264 |
Five-minute samples from live/summary.json, 27 August 2026. The Extended Look
premiered on Netflix at 19:00 UTC (3 p.m. ET) and
ran 26 minutes.
The audience began moving at 18:35, twenty-five minutes before the premiere, and peaked at 19:25 — minute twenty-five of a twenty-six-minute show. By 19:55 it had given back two thirds of the rise. The curve tracks the runtime almost exactly, which is the strongest evidence we have that what we are looking at really is the Extended Look and not a coincidence.
Against the ordinary daily rhythm, the size of it is easier to see hour by hour:
| Hour (UTC) | Actual | 28-day mean for that hour | σ |
|---|---|---|---|
| 27 Aug 18:00 | 420,048 | 314,148 | +2.7 |
| 27 Aug 19:00 | 824,338 | 338,978 | +13.8 |
| 27 Aug 20:00 | 540,755 | 382,730 | +3.3 |
| 27 Aug 21:00 | 508,387 | 422,159 | +1.4 |
| 27 Aug 22:00 | 439,997 | 408,886 | +0.5 |
| 28 Aug 00:00 | 328,784 | 325,230 | +0.1 |
| 28 Aug 01:00 | 270,173 | 299,322 | −0.5 |
| 28 Aug 02:00 | 339,163 | 267,873 | +1.2 |
Hourly means of five-minute samples, against the mean for the same hour across the 28 complete UTC days ending 26 August (n = 28–29 per hour). The 19:00 hour is almost fourteen standard deviations above its own normal. Four hours later the category was back inside one.
It was YouTube, and only YouTube
The interesting part is not that GTA streaming spiked. It is where.
YouTube’s live GTA audience went from 71,129 at 18:00 to 706,397 at 19:25 — 9.9× in eighty-five minutes, and 57.6× its median for the day. At the peak, YouTube held 73.7% of all live GTA viewing on the planet. On an ordinary day it holds somewhere between 1% and 15%, a range we measured two days ago and which was already the most volatile thing in our data.
Kick, which normally carries the majority of this category, barely participated. It was at 172,117 before the show, 175,662 at the peak — below where it started — and did not reach its evening high until 21:30, long after the reveal was over. The Kick audience is predominantly Spanish, Portuguese and Arabic-speaking, and it appears to have carried on with its Thursday night largely undisturbed.
Twitch went the wrong way
Twitch had 96,238 live GTA viewers at 18:00. Ninety minutes later, in the middle of the biggest Grand Theft Auto moment of the year, it had 77,019. By 20:30 it had 44,143 — a 54% fall from where it started, while Kick over the same window rose 42%.
That is because Twitch broke. Downdetector logged more than 17,000 complaints at the peak, Twitch’s own status page flagged a major outage across web, chat and video viewing, and Kai Cenat’s and IShowSpeed’s profiles were unavailable as of 4:45 p.m. ET (Forbes, 27 August). Twitch acknowledged the problem at around 4 p.m. ET — 20:00 UTC.
Because Kick and YouTube were running normally in the same minutes, the ratio between them is a cleaner instrument than either number alone. Twitch’s live GTA audience divided by Kick’s:
| Window (UTC) | Twitch ÷ Kick |
|---|---|
| 16:00–17:50, before the show | 0.595–0.690 |
| 19:00–20:10, during and just after | 0.279–0.343 |
| 20:20–20:40, outage acknowledged | 0.181–0.220 |
| 21:00–21:50, after recovery | 0.279–0.308 |
The ratio held a tight band all afternoon, stepped down when the co-stream audience moved to YouTube, dropped again to a distinct low at 20:30, and settled back at the post-event level within twenty minutes. The step at 18:20 is the audience choosing YouTube. The dip at 20:30 is the outage, and it is the only part of the day that comes back.
Two limitations, and they cut in the same direction. We only count streams categorised as Grand Theft Auto, and the reported 881,000 figure against our own peak of 81,200 on Twitch is a strong hint that many of the largest Twitch reaction streams were categorised as something else — a creator reacting to a Netflix show is often sitting in Just Chatting. So our Twitch number is not the Twitch co-stream audience, and the outage’s real cost was much larger than the ~24,000 viewers we can see it take. And we count concurrent viewers, not attention: an unwatched tab counts the same as a rapt one.
The free release did nothing
This is the finding we did not expect, and it contradicts something we published three weeks ago.
Rockstar put the Extended Look on its own YouTube channel and the GTA VI site at 9 p.m. ET — 01:00 UTC on 28 August — free, no subscription, worldwide. On 7 August we argued that the six-hour delay would not cost Rockstar anything, because 01:00 UTC historically averages 5.2% more live GTA viewers than 19:00 does. On the hour-of-day curve, the free slot was the better slot.
It wasn’t close. The 19:00 hour ran at 824,338 and the 01:00 hour at 270,173 — the paywalled premiere beat the free release by 3.05×. And 270,173 is not merely lower; it is below average for that hour, half a standard deviation under the 28-day mean of 299,322. Minute by minute, the category was falling before 01:00 and kept falling through it: 330,131 at 00:55, 318,691 at 01:00, 234,581 by 01:55. There is no step anywhere in the window. YouTube’s own live GTA audience went from 22,267 to 12,560 across the same hour.
The one bump after the free release runs from 02:00 to 02:50, and it is Kick, not YouTube: Kick went from 172,749 to 351,484 while YouTube kept sliding, 11,912 to 9,417, across the very same minutes. The 02:00 hour finished at +1.2σ, comfortably inside its own normal range, and the channels sitting at the top of Kick through that part of the night are Spanish- and Portuguese-language freeroam streams. Whatever it was, it was not the Extended Look — the video had been free for an hour and YouTube was at its lowest point of the day.
We were wrong for a reason worth stating, because it generalises. The hour-of-day curve measures when people are available to watch. It cannot see that a co-stream audience is manufactured by scarcity. At 19:00 the only free way to see the reveal at the moment it happened was to watch it over a creator’s shoulder; six hours later you could simply press play. The exclusivity did not suppress the audience — it created the event.
Scoring what we said
We publish falsifiable claims and grade them. Three came due.
| Claim | Where | Result |
|---|---|---|
| 19:00 UTC 27 Aug exceeds 403,363 concurrent | 23 Aug | Correct — 824,338, 2.04× the threshold |
| 01:00 UTC 28 Aug exceeds 343,728 concurrent | 23 Aug | Wrong — 270,173, below the hour’s own average |
| No writer will describe playing GTA 6 | 27 Aug | Holds so far — the previews that went up with the premiere are hands-off throughout |
One right, one wrong, one standing. The one we got wrong is the one that taught us something.
A fourth claim was left open here, and the window has since closed.
Scored 29 August: we were wrong. On 26 August we predicted that at least three major outlets would compare the Extended Look’s YouTube view count to Trailer 1’s 90,421,491 without mentioning that YouTube changed how views are counted on 24 August. The 24-hour window closed at 01:00 UTC on 29 August, and we could not find a single outlet that made the comparison — never mind three.
The reason we got it wrong is more interesting than the prediction was. We assumed the Extended Look would be read as a record-chasing trailer. It was not, because it did not perform like one: Gameranx reported 4 million YouTube views on 28 August, against a trailer record more than twenty times that. Nobody compares a number to a record it is nowhere near. A 26-minute gameplay showcase that premiered on Netflix six hours earlier, and that most of its audience watched over a creator’s shoulder, was never going to post a trailer’s first-day count — which is the same finding as the rest of this post, arriving from the other direction.
The asterisk from 26 August still stands for the next actual trailer. It simply had nothing to attach itself to this time. One caveat on the grading: we searched, we did not audit every outlet on the internet, so read this as “we could not find it” rather than “it does not exist”.
This morning
The category has already normalised. Across 14 complete censuses between 04:06:38 and 04:20:38 UTC on 28 August, live GTA viewing averaged 231,606 concurrent viewers on 1,984 channels:
| Platform | Viewers | Share | Channels |
|---|---|---|---|
| Kick | 193,807 | 83.7% | 296 |
| Twitch | 32,380 | 14.0% | 1,590 |
| YouTube | 5,419 | 2.3% | 99 |
| All three | 231,606 | 100% | 1,984 |
Mean of 14 censuses, 04:06:38–04:20:38 UTC, 28 August 2026. Board and tail fetched
together with t matched in all 14; viewer and channel sums reconcile to
summary.json exactly in all 14; zero duplicate (platform, stream) pairs; no
degraded platforms.
YouTube is back to 2.3% from 73.7%, on 99 channels. Kick’s 83.7% is high even for Kick — it ran at 75.7% in yesterday’s census at the same hour — but that is the ordinary overnight pattern, not an aftershock.
The record stood for about ninety minutes.
What this does not show
Our sampling is a concurrent-viewer census taken once a minute, so a spike shorter than a minute is invisible and a five-minute figure is a snapshot, not an average. We cannot attribute the peak to individual channels — we hold a category-level series, not a per-channel history — so we cannot tell you who the 706,397 were watching. Freeroam remains our fallback classification, so the content-type split during the event is unreliable and we have not quoted it. And, most importantly: none of this measures how many people watched the Extended Look. It measures how many watched it with company.
Cite this. Peak concurrent live Grand Theft Auto viewers across Kick, Twitch
and YouTube: 959,078 at 19:25 UTC on 27 August 2026; the 19:00 hourly mean was
824,338 against a 28-day mean of 338,978 for that hour. Free-release hour, 01:00 UTC
28 August: 270,173. Permalink:
https://luciastream.com/blog/how-many-watched-gta-6-extended-look/. Sampling
window and method: five-minute resolution from a once-a-minute census of every live
GTA stream on the three platforms, described in full on our
methodology page. The current figure is
always on the live dashboard.
The Extended Look itself, on Rockstar’s channel:
Numbers in this post come from our own sampling — seehow we measure.