YouTube Statistics, With Sources

Every number below carries the source that stated it and the date we last checked that source. There are fewer numbers here than on most YouTube statistics pages, and that is the point: this list contains only figures we could trace to a primary source and verify ourselves.

If you are here to cite a number, the citation you want is the primary source. We link it every time. What this page adds is knowing which widely-repeated figures no longer trace to anything.

What YouTube currently states

These four figures are the ones YouTube publishes on its own press page. Checked 31 July 2026.

Four figures is a short list. Most pages titled "YouTube statistics" carry fifty, and the difference is not research effort — it is that they publish numbers whose origin they have not checked.

Arithmetic on those numbers

These are ours, derived from the figures above. The derivation is stated so you can check it rather than trust it.

• Roughly 231 videos uploaded every second. 20 million ÷ 86,400 seconds in a day. If you are reading this sentence at an average pace, about forty videos went up while you did.
• Roughly 2.3 million Shorts views every second. 200 billion ÷ 86,400. Writing this down caught an error in our own first draft, where that figure had been attached to the wrong denominator and came out 870 times too small — which is the argument for showing the arithmetic rather than the argument against it.
• Around 7.3 billion videos a year, at the current daily rate.

Where a page gives you a number like this without showing the arithmetic, you cannot tell whether it was derived or invented.

Widely quoted, no longer traceable

This is the section other statistics pages do not have, and it is the reason this one exists.

"500 hours of video uploaded every minute"

The most-quoted YouTube upload statistic in existence. It is not on YouTube's press page as of the date above. YouTube's current published figure is in videos per day, not hours per minute, and the two cannot be converted into each other without knowing average video length — which nobody publishes.

So when an article states 500 hours per minute alongside a 2026 date, it is repeating a figure from an older source without saying so. The number may well still be roughly true. The citation is what is broken.

"2 billion monthly logged-in users"

Widely repeated, and not among the figures YouTube currently publishes on its press page. Later statements have used higher numbers. If you need a user count for something that matters, take it from a dated primary source and cite the date, because this figure has moved several times and the internet's copy of it has not.

"X million YouTube channels"

Channel-count figures circulate widely and YouTube does not publish one. Every number you have seen is an estimate produced by a third party using a sampling method, and the methods differ enough that the estimates differ by large multiples. An estimate is not wrong for being an estimate — but it should be labelled as one, with the method named, and it almost never is.

How this page is maintained

Three rules, and they are the whole method:

• A number gets published only if we can name the primary source and read it ourselves. Second-hand citations of second-hand citations are how a 2019 figure ends up in a 2026 article.
• Every number carries the date we last checked it. A statistic without a date is not a statistic, it is a rumour with digits.
• Derived figures say they are derived, and show the arithmetic.

The corollary is that this page will always be shorter than its competitors, and it will sometimes say "we do not know". Both are deliberate.

Numbers we can verify ourselves

Some facts about YouTube are not statistics from a press release — they are behaviours of the platform that can be checked directly through the official API, and we build tools that do exactly that. These are things we can state from our own observation:

• Subscriber counts above 1,000 are rounded to three significant figures — measured across 559 channels, with no exceptions. See the section below; this is our own measurement rather than a restatement of the documentation.
• The public dislike count returns zero. Removed in November 2021 and never restored to the API. Any current dislike figure is a pre-2021 archive or an extrapolation from browser-extension users.
• Dead videos in playlists are rarer than commonly claimed. We sampled 873 items across 25 playlists surfaced for “full course tutorial” on 31 July 2026 and found zero deleted or private entries. This corrects a figure we ourselves published earlier, which asserted around ten per cent without measuring it. Sample caveat: these are playlists YouTube surfaced for one query, not a random sample of all playlists.
• Comment threads return 100 per API request at one quota unit each. A video with 80,000 comments costs 800 units to read once, which is why no tool searches all the comments on a large video.

Measured: subscriber rounding across 559 channels

YouTube documents that subscriber counts are abbreviated above 1,000. What it does not publish is how strictly that holds, or what the resulting uncertainty is at each channel size. We measured it on 31 July 2026.

Method. 574 channel IDs collected from trending videos across 14 regions, plus two channel searches to reach smaller channels. Statistics read through the official API in batches. 15 channels hide their subscriber count and were excluded, leaving 559. For each, we tested whether the published count is an exact multiple of 10^(digits−3) — the step implied by three significant figures.

Result: 499 of 499 channels above 1,000 subscribers sat exactly on their step. No exceptions. The rule is not a tendency or an approximation — in this sample it is absolute.

Below 1,000 the counts are genuinely exact rather than rounded to something smaller. Of 60 channels sampled under the threshold, 56 were not multiples of 10 — values like 739, 237 and 3. A system rounding to the nearest ten could not produce those.

What follows from it

• A live subscriber counter for a 10M+ channel is reading a number that moves in steps of 100,000. Fifty thousand people can subscribe without the display changing. Any counter animating single digits several times a second is generating every digit between the readings it actually has.
• Milestone countdowns for large channels are uncertain by ±50,000. A channel shown as 40,000 short of 10 million could be anywhere from 90,000 short to already past it.
• Small channels are the only ones where a live counter is literally accurate, and they are the ones nobody builds counters for.

Sample caveat. Trending skews toward large and active channels, and the two searches used to reach smaller ones return whatever YouTube surfaces for those queries. This is not a random sample of all YouTube channels, and the band sizes above reflect that: 234 of the 559 are over a million subscribers. What the sample does support is the absence of counterexamples — 499 chances for the rule to break, and it did not break once.

You can check any single channel yourself with the Live Subscriber Counter, which prints the step size next to the count.

Measured: how often a view count actually changes

The second thing we can observe rather than quote. Measured 31 July 2026.

Method. The 40 top trending videos in the US, polled every 25 seconds for 7.1 minutes through the official API, bypassing our own cache. 86 changes recorded across the set.

The busiest video in the sample — a trailer at 24 million views — changed at every single poll, meaning its true cadence is faster than 25 seconds and our sampling could not resolve it. The 25-second figure in the range is a floor imposed by the method, not a property of YouTube.

What this does and does not tell you

It does not measure lag. How far the published count sits behind the true one would require the true count, which exists only inside YouTube. We previously described the counter as trailing reality "by minutes" on our own view counter page; that was an assertion, not a measurement, and it has been removed. What can be measured is the cadence above.

Zero decreases in seven minutes does not contradict the well-known fact that view counts fall when invalid traffic is filtered. Seven minutes is simply far too short a window to observe it — the filtering happens over hours and days. It is listed as observed-zero rather than absent.

A practical consequence: a live view counter polling once a minute is sampling roughly twice per change at the median, which is adequate. For the fastest-moving videos it is under-sampling, and no polling interval short of continuous would fix that.

Measured: what a normal like rate looks like

Third measurement, and the one that most changed what we had written. Measured 31 July 2026.

Method. 120 channels collected from trending across 10 regions, then ten videos from each channel's uploads playlist rather than the trending videos themselves — a channel's back catalogue is closer to a normal video than its one viral hit. 1,152 videos with a public like count and at least 100 views.

The commonly quoted "2–5% is typical" describes only 43% of videos. We published that band ourselves. In this sample 438 videos fall below 2% and 215 above 5% — so a creator at 1.5% who reads that band concludes something is wrong, when they are sitting close to the 25th percentile of a normal distribution.

Category matters more than size

A 4.4× spread between People & Blogs and Music. Comparing your like rate against a single global figure, without accounting for what you make, is comparing against the wrong population.

And it does not rise with audience size

The peak is in the middle, not at the top. Videos in the 10K–100K band have the highest median like rate in the sample, and million-view videos are close to the bottom — plausibly because reaching a million views means reaching people who were not looking for you, and those people watch without voting. We are reporting the shape, not explaining it.

Sample caveat. Every channel here reached trending in some region, so this is a sample of comparatively successful channels' back catalogues, not of YouTube as a whole. Small independent channels are under-represented in the lower view bands. Also worth noting: none of the 1,152 videos hid its like count, so hiding it appears to be rare among these channels.

FAQ

Why does this page have so few statistics?

Because it only contains numbers we traced to a primary source and checked ourselves. Pages with fifty statistics have not done that for fifty numbers; they have copied a list. Ours grows when we verify something new.

Can I cite these numbers?

Cite the primary source we link, not us — that is what a source is for. If you want to cite the analysis of which figures are stale, cite this page and note the date, because the situation changes when YouTube publishes something new.

How many videos are uploaded to YouTube every day?

Over 20 million, per YouTube's own press page. That works out at roughly 231 a second. Note the unit: YouTube publishes videos per day, not the hours-per-minute figure that circulates more widely.

How many views do YouTube Shorts get?

Over 200 billion a day, averaged, per YouTube's press page. For context on what that means for a creator, Shorts views count toward a separate monetization threshold — see monetization requirements.

How accurate is a live subscriber counter?

Exactly accurate below 1,000 subscribers, and progressively less so above it. We measured 559 channels and every one above the threshold sat exactly on its three-significant-figure step — meaning a 10M+ channel is published in increments of 100,000, and the true figure can be 50,000 either side. See the measurement above.

Is "500 hours per minute" wrong?

Not necessarily wrong — untraceable. It is not among the figures YouTube currently publishes, and it cannot be derived from the ones that are. Use it if you can find a dated primary source for it; do not use it with a 2026 date attached simply because everyone else does.

Next steps

• Check any channel's real public numbers with Channel Statistics.
• See the subscriber rounding for yourself with the Live Subscriber Counter.
• Read what the API does and does not expose about a single video in the Metadata & Rights Viewer.