Did YouTube change how views are counted? Yes, on 24 August

Yes. Since 24 August 2026, YouTube counts a view the moment a video starts playing — no minimum watch time, and not only on Shorts. Google’s own help page puts it in one sentence:

Beginning August 24, 2026, views are counted the moment a video starts to play across all formats, including Shorts, long-form videos (VOD), and live streams.

That is the whole change, and it is three days old. Everything below is what it does to numbers you are about to see quoted.

The short version

Before, a long-form video had to be watched past some threshold before the counter moved. YouTube never published what that threshold was; the figure reported everywhere is about thirty seconds, and it comes from outlets rather than from Google — LiveReacting’s write-up on 18 August is representative: “until now, a long-form video needed roughly 30 seconds of watch time before YouTube counted a view.” Treat the exact number as reported, not confirmed. The direction is not in doubt.

Now the counter moves on the first frame. On a live stream, the same source says, “a view will count the moment someone enters the broadcast” — which is a meaningfully different event from someone watching your broadcast.

Two things did not change, and both matter more than the headline.

Money did not change. The help page is explicit: “This update won’t impact your YouTube Partner Program (YPP) earnings or eligibility. YPP earnings will still be based on ‘engaged views’ and ‘engaged watch hours’. YPP eligibility will still be based on ‘qualified views.’” So there are now two numbers — a public one that got easier to earn and a private one that did not. Forbes covered it that way on 17 August, and that framing is the right one.

Old numbers did not change either. Existing videos keep the totals they already had; the new method applies only to plays they collect from 24 August onward. That is reported rather than stated on the help page, but nobody disputes it, and it is the detail that breaks everything in the next section.

Which brings us to tomorrow

Grand Theft Auto VI: An Extended Look premieres on Netflix at 3 p.m. ET on Thursday 27 August, and lands free on Rockstar’s YouTube channel and the GTA VI site at 9 p.m. ET the same day. That is Rockstar’s own wording, from the Newswire post.

Within about ninety minutes of that YouTube upload, someone will publish a piece comparing its view count to the record. The record is GTA VI Trailer 1, which took 90,421,491 views in 24 hours in December 2023 and holds a Guinness World Record for it. We wrote about whether the Extended Look can beat it on 23 August, and argued it probably cannot, because a six-hour Netflix exclusive drains the concentration a 24-hour record is really measuring.

Here is the part that post did not know, because the change had not happened yet: that comparison is no longer apples to apples. Trailer 1’s 90,421,491 was accumulated under the old rule and has not been restated. The Extended Look’s number will be accumulated under the new one, from the first frame. Whatever the two numbers turn out to be, one of them was measured with a ruler that has since been re-marked.

I want to be careful about what that does and does not mean. It does not mean Rockstar gets a free record, or that any figure published tomorrow is fake. The direction of the bias is obvious — first-frame counting can only raise a count, never lower it — but the size is unknown and unknowable from outside, and on a video this anticipated the share of people who bail in the first thirty seconds may be genuinely tiny. It might be worth a percent. It might be worth fifteen. Nobody outside YouTube can say, which is exactly why the comparison should carry an asterisk rather than a correction.

Streams Charts, who cover this beat properly, made the general form of the point on 25 August: “similar View totals can represent very different levels of audience consumption.” As far as I can find, nobody has yet connected it to the biggest video launch of the quarter, which happens tomorrow night.

What we can measure, and what we can’t

We track every live GTA stream on Kick, Twitch and YouTube once a minute. We record concurrent viewers — how many people are watching at a given second — which is a different metric from views and one the 24 August change does not touch. We do not collect YouTube’s view counts at all, so we cannot measure the inflation directly. Anyone who tells you they can size it from outside is guessing.

What we can do is describe the platform the change lands on. Here is live GTA this morning:

Platform	Viewers	Share	Channels	Viewers per channel
Kick	226,878	69.9%	327	693.8
Twitch	91,432	28.2%	1,611	56.8
YouTube	6,248	1.9%	112	55.8
All three	324,558	100%	2,051	158.2

Mean of 15 complete censuses, 04:07:52–04:22:54 UTC on 26 August 2026. Board and tail fetched together with the publish timestamp matched across all three files; viewer and channel sums reconcile to the platform summary exactly in all 15; zero duplicate (platform, stream) pairs; no degraded platforms. YouTube’s share held between 1.78% and 2.16% across the window.

The limitation to state before anything is built on that 1.9%: it is one twenty-minute window at one of the quietest hours of the global day, and YouTube’s GTA number is the least stable of the three by a wide margin.

YouTube’s number is one broadcast wide

This is the finding worth taking away, and it is ours because it comes from our own committed history. Every deployment writes a dated digest into the repository, so we can reconstruct what the platform split said at sixteen separate moments between 19 and 25 August, and add this morning’s census as a seventeenth.

YouTube’s share of all live GTA viewing across those seventeen readings runs from 1.02% to 15.26% — a factor of fifteen. Some of that is the clock: we found in August that YouTube peaks and troughs at different hours from Kick and Twitch, so its share is highest when the other two are quietest. But the clock does not explain this pair:

Reading	YouTube viewers	YouTube channels	Share of all live GTA
03:43:10 UTC, 24 Aug	21,016	119	9.50%
03:53:41 UTC, 25 Aug	2,705	80	1.31%
04:07–04:22 UTC, 26 Aug	6,248	112	1.92%

Ten minutes apart on the clock, one day apart on the calendar, and a 7.3× difference in YouTube’s share. The channel count barely moved. That is one large broadcast being live or not being live.

At this morning’s reading the single biggest GTA channel on YouTube held 15.4% of YouTube’s entire live GTA audience, and the top three held 38.7%. Concentration like that is why a single number for “YouTube’s share of GTA” is close to meaningless unless you say which minute you took it in — and it is why, from tomorrow, a views-based version of the same claim will be worse rather than better, because the metric underneath it just moved too.

One more thing the census turned up

Between six and seven of the roughly 112 live GTA channels on YouTube this morning were broadcasting under the same title — the same words, the same emoji, varying only in hashtags. Between them they held 26.4% to 33.0% of YouTube’s entire live GTA audience across the fifteen censuses, averaging 28.6%.

There is nothing improper in that and I am not suggesting otherwise: recycling a title and thumbnail that works is ordinary practice, it breaks no rule, and I have no idea whether those channels are related to each other. It is worth noting only because of what changed on Monday. A metric that counts a view the instant playback begins is, mechanically, most generous to formats built around the click rather than the watch — and roughly three in ten of YouTube’s live GTA viewers this morning were already inside one.

By contrast, zero of Kick’s 327 live GTA channels shared a title with another, and 28 of Twitch’s 1,611 did, most of those advertising the same Rockstar Drops campaign.

So what should you actually do with a view count now

Three rules, and they are boring on purpose.

Do not compare a video’s 24-hour count to a pre-24-August record without saying so. Trailer 1’s 90,421,491 is a real number, certified, and it is now a historical artefact measured differently from anything published this week. Even that figure is contested, incidentally — YouTube’s own blog said “over 93 million” for the same event that Guinness certified at 90,421,491, and any “did it beat the record” headline tomorrow will quietly pick one.

Do not read a step change in your own channel’s views as an audience change. If your numbers jumped this week and your watch time did not, nothing happened to your channel; the definition moved underneath it.

And for cross-platform comparison — which is the thing this site exists to do — use concurrent viewers, not views. Not because concurrency is a purer metric, it is not: it undercounts everything asynchronous and it says nothing about attention. It is just the one metric that means the same thing on Kick, on Twitch and on YouTube, and still means the same thing this week as it did last week. You can see the current three-platform picture on our live dashboard and the 30-day creator ranking on the leaderboard.

The ledger

Claim	Status
Views counted from the first frame across Shorts, VOD and live streams from 24 Aug 2026	Confirmed — YouTube Help, How video views are counted
YPP earnings and eligibility unaffected; still “engaged views”, “engaged watch hours”, “qualified views”	Confirmed — same page
The old long-form threshold was roughly 30 seconds	Reported — LiveReacting and others; YouTube never published the figure
Pre-24 August totals are not restated; the new method applies to plays from 24 Aug onward	Reported — LiveReacting, 18 Aug
A live-stream view counts the moment a viewer enters the broadcast	Reported — LiveReacting, 18 Aug
Extended Look: Netflix 3 p.m. ET 27 Aug, YouTube and the GTA VI site 9 p.m. ET 27 Aug	Confirmed — Rockstar Newswire
Trailer 1: 90,421,491 YouTube views in 24 hours	Confirmed — Guinness World Records
Trailer 1: “over 93 million” in 24 hours	Reported — YouTube’s official blog, same event, different figure
Trailer 2: ~69m YouTube views in 24 hours, 475m+ across all platforms	Reported — RockstarINTEL and Rockstar via GameRant
The size of the inflation on any given video	Unknown — not measurable from outside YouTube

Citing this

The measurements above are ours and you are welcome to use them with attribution to luciastream, permalink https://luciastream.com/blog/did-youtube-change-how-views-are-counted/. The sampling window is 04:07:52–04:22:54 UTC on 26 August 2026, 15 complete censuses of every live Grand Theft Auto stream on Kick, Twitch and YouTube; the seventeen-reading series runs 19–26 August 2026. How the collector works, what it can and cannot see, and where the classification is weakest are all written up on our methodology page.

One prediction to score, since the whole point of the piece is that numbers need to be checkable. The Extended Look’s YouTube upload will be reported against Trailer 1’s 90,421,491 by at least three major outlets within 24 hours of going live, and none of them will mention that the counting rule changed on 24 August. If I am wrong, I would rather be wrong about this than about most things.