48% Make One AI Video and Stop: What 2,224 Videos Show

Most advice about faceless channels is written from the outside: public CPM tables, "typical creator" ranges, and the same five tips rewritten. We can do something different, because we can look at what people actually did. Every figure below was read from our production database on 3 October 2026, over the 90 days before it. Nothing is estimated, and where a number does not exist we say so instead of filling the gap.

Some of it does not flatter us. That part is included, because a post that only reports the good numbers is just an ad with a chart in it.

What 90 Days Looks Like

Over the 90 days to 3 October 2026, on AutoShort:

• 2,297 videos were started and 2,224 finished — a 96.8% completion rate
• 788 people started at least one; 766 finished at least one
• 47 accounts have ever connected a YouTube or TikTok channel, across the whole life of the product

The completion rate is the part we are happy with. The rest of this post is about what happened after the video existed, which turns out to be the harder problem.

Half Make One Video and Stop

Finished videos per creator, same 90-day window:

• 367 creators finished exactly one video — 48% of everyone who finished anything
• 226 finished two
• 136 finished three to five
• 27 finished six to twenty
• 10 creators finished twenty-one or more — and between them produced 731 videos, a third of everything made in the window

Ten people out of 766 — 1.3% — account for a third of the output. The distribution is not a bell curve with a long tail; it is a cliff at one video, and then almost nothing until you reach a handful of people treating this as a daily job.

It is worth being precise about what stopping at one video means. It does not mean the video failed: 96.8% of videos finish. It means the person made the thing they came to make, looked at it, and did not come back. Whatever the second video costs them, it is not credits.

The One Setting That Separates Them

We split the same 766 creators by a single fact: whether they had ever connected a YouTube or TikTok channel to their account.

• Not connected — 744 creators, averaging 2.4 finished videos
• Connected — 22 creators, averaging 21.4 finished videos

That is 8.9 times more output, and it is the largest difference we found between any two groups of our users.

It is a correlation, and we are not going to dress it up as a cause. Someone who connects a channel on day one is, by definition, someone who already intends to publish. The connection may be producing the consistency, or the intent may be producing both. One measurement cannot tell the two apart, and we have not run the experiment that would.

What we can say is narrower and still useful: in our data, the people who got as far as connecting a destination are the people still making videos three months later, and the people who did not are overwhelmingly the people who made one.

What We Cannot See

Of the 2,224 finished videos, 10 were ever scheduled to a platform and 6 were actually posted.

Read that carefully, because the obvious reading is wrong. That table only records videos published through us. A creator who downloads an MP4 and uploads it to TikTok by hand leaves no trace in it at all. So the honest statement is: six videos were auto-posted by us, and we have no idea how many were posted by hand. It is a measure of our publishing feature, not of our creators.

We are stating the number anyway, with its limit attached, because the limit is itself the finding: for a product whose pitch is "set it and forget it", almost nobody is using the part that does the forgetting. 47 accounts have ever connected a channel. That is the number we would most like to be wrong about, and it is not.

Why Publishing Is the Bottleneck

Put the three measurements next to each other and the shape is hard to miss. Finishing a video is close to solved — 96.8%. Making a second one is where half the people leave. And the group that does not leave is the group that wired up somewhere for the videos to go.

This matches the one customer whose channel we looked at end to end in our two-week case study: 42 videos made, 24 published, 8,689 views. Our own five demo channels hold 43–56 videos each and average 11–26 views. The difference between those outcomes is not the tool — it is the same tool — and it is not the number of videos made, which is comparable. It is that one of them was publishing to a real channel in a real niche and the others were sitting in a library.

A video nobody can watch is worth exactly nothing, and it costs the same credits as one that gets 1,217 views.

What To Do With This

If you are starting a faceless channel this week, the data says one thing loudly:

• Connect the destination before you make the first video. It takes a minute, and it is the only setting that separated the creators who kept going from the ones who did not.
• Plan for the second video, not the first. The first one is easy enough that 766 people made one. Deciding what the channel is about is the thing that makes a second one obvious.
• Pick a niche narrow enough to be repeatable. The channel in our case study posts one astronomy fact per video. That format can be made 42 times without thinking about it, which is exactly why it got made 42 times.

What We Will Not Claim

Things this data does not show, which we are not going to imply:

• That connecting a channel causes consistency. It correlates with it, 8.9x, measured once. We have not isolated the cause.
• That 48% quitting is unusual. We have no comparable figure from any other tool, so we cannot tell you whether that is bad, normal, or good for this category.
• That our creators only published six videos. Six is what we posted for them. Manual uploads are invisible to us and we are not going to pretend otherwise.
• Any revenue figure. We do not measure what these channels earn, so we are not going to quote a number for it.

These numbers are one snapshot, taken on 3 October 2026. A single measurement is a number; a series of them is a trend. We will publish the trend when we have one, including if it is worse.

AutoShort's free tier is 100 credits — about two videos — with no card required. If you take one thing from this post, take the first bullet above and connect your channel first.