What we run

YouTube growth, run for you.

YouTube is the slowest and most production-heavy channel we run. Research, scripting, titles, and descriptions scale well with an AI layer. Appearing on camera does not. Neither does the edit, or the conversation in the comments and community tab. Zway builds everything up to the record button, then a person handles the rest.

Last reviewed 27 August 2026

The failure mode

What breaks when AI does this alone.

A model can write the script. It cannot be the person on camera, and on YouTube the person is most of the value. It also cannot make the edit decisions that hold attention, because those come from watching the footage back and knowing which twelve seconds are boring. Then there is the comment section and the community tab, which is a live conversation with people who actually watched. An unattended account either ignores them or answers generically, and both read as abandonment.

What our operators do here

Turn the research into a script a founder can read out loud without sounding like a document
Run the recording session, including the retakes, so nobody is directing themselves
Cut the video, choose the thumbnail, and defend the first thirty seconds
Write titles and descriptions that match what the video actually delivers
Answer comments and community tab posts as a named person, within the day
Kill video ideas that will not survive the effort, before anyone books a recording
Report retention curves and say plainly which videos are not working

Volume

What ships in a typical week.

Typical weekly output on the YouTube channel
What	Typical volume
Scripts drafted and edited to camera-ready	1 to 2
Long-form videos published	2 to 4 per month
Shorts cut from footage you already recorded	3 to 5
Comments and community replies handled	All of them
Video ideas killed before production	Logged and reported

Volumes are typical rather than contractual. They move with what the channel is actually returning.

What good looks like

The metric that matters

Average view duration climbing video over video, and comments that ask real questions

Realistic timeline

Four to six months before the back catalogue does any work for you. The first videos will be bad.

YouTube is the slowest channel we run, and we would rather say so first

If you need pipeline this quarter, do not start here. YouTube is the most expensive channel to produce and the last one to pay you back.

We still run it, for two reasons. A video keeps working after it is published, which is the opposite of paid, where the traffic ends when the budget does. And a founder explaining their own product on camera is the most convincing asset a small company can own, because it cannot be faked by a competitor with a bigger budget.

Both of those arguments are about years, not weeks. If that is not the horizon you are working on, pick a different channel and come back.

What scales here, and what does not

The split is unusually clean on this channel, which is why it is a useful one to look at closely.

Part of the work	Handled by the AI layer	Needs a person
Topic research and demand checks	Yes, at volume	A sanity check
Script drafting	Yes, first and second pass	Rewrite for how you actually talk
Titles, descriptions, chapters	Yes	Final call on the promise being made
Being on camera	No	You, or someone at the company
The edit	No	Watching it back and cutting
Thumbnail selection	Concepts only	The choice
Comments and community tab	Drafts only	Every reply that goes out

Everything in the left column is a research and language problem, and it is genuinely fast now. Everything in the right column is a judgment or a presence problem, and no amount of model quality changes that.

The record button is the boundary

The script is not the hard part. Reading it without sounding like someone reading is the hard part, and so is deciding, twenty minutes into a recording, that the framing is wrong and starting again.

We run the session so that you are not directing yourself. An operator watches the take, calls the retake, and keeps a list of the moments worth keeping. That is a small thing, and it is the difference between a founder recording once and never again, and a founder who ships two videos a month for a year.

Then the edit. Retention is decided in the cut, not the script, and the cut is made by someone who watched the raw footage and knows which twelve seconds are boring. A model has not watched your footage. It cannot tell you that the third minute drags.

The comment section is the channel, not the afterthought

The people in the comments watched the whole thing. They are the closest thing to a warm lead this channel produces, and most companies ignore them.

An operator answers them as a named person, within the day, including the ones that are wrong about your product and the ones that are rude. They also run the community tab, which is where you find out what to make next without guessing, because people tell you directly.

Generic replies are worse than none. Everyone reading can tell, and the thread stays up permanently.

When we tell you not to run this

We will say no to YouTube when there is no product to demonstrate yet, when nobody at the company is willing to appear, or when the message has not been tested anywhere cheaper first. Video is the most expensive place to discover that your positioning is wrong.

We will also kill individual video ideas before anyone books a recording. A weak idea costs the same to produce as a strong one, and the production cost is the whole problem with this channel.

YouTube questions

Can AI just make the videos for us?

It can do everything up to the record button. Topic research, scripts, titles, descriptions, chapters, and the first draft of a thumbnail concept all come out of the AI layer. Synthetic presenters exist and we do not use them for founder-led content, because the entire reason this channel works is that a viewer believes a real person built the thing and is explaining it.

Do I have to be on camera?

For most of what we recommend, yes, or someone at the company does. There are formats that avoid it, screen recordings, product walkthroughs, and narrated demos with a voiceover, and they work for technical content. What they do not build is the recognition that makes someone click the next video because of who is in it. We will tell you which trade you are making.

How long before YouTube produces signups?

Four to six months is the honest range before the channel contributes anything you would notice in a pipeline review. Individual videos can land earlier by accident. The reason we still run it is that the videos keep working after they are published, which is the opposite of paid, where the traffic stops the day the budget does.

Is YouTube worth it for a pre-launch startup?

Usually not as the first channel. Before there is a product to demonstrate and a message that has already survived contact with real people, you are spending your most expensive production hours guessing. We would rather prove the message somewhere cheap and fast first, then bring the winning material to video.

Who owns the channel, the footage, and the edits?

You do. The channel is yours, the raw footage is yours, and the project files are yours. Operators work inside your account rather than a shared one we control, which is the same arrangement we use on every channel. If you stop working with us, nothing has to be handed back because nothing was ever held.

See what we would run on YouTube.

Thirty minutes. You leave with the thirty-day plan either way.

Book a demo

30 minutes. You leave with the plan either way.