A scripted series. YouTube's inauthentic content policy targets video that's generic, repetitive, and has no narrative arc — its own trust and safety lead's words. A short drama is the structural opposite: recurring characters, continuity between episodes, and one forward move per episode. The format that's hardest to mass-produce is the format the policy protects.
Most coverage of YouTube's July 2026 update read it as a threat. Read it as a brief instead and it tells you what to make.
What is YouTube actually describing?
Its own framing is more useful than any of the coverage.
Matt Halprin, YouTube's Vice President of Trust & Safety, made the point in both directions in the same breath. The new tools let creators produce a higher volume of good work, which YouTube wants in the Partner Program. The same tools also let someone make lots of videos quickly that are very similar — and those, he said, are:
very generic and don't really have a narrative arc
Matt Halprin, VP of Trust & Safety, YouTube — Creator Insider, July 2026
He added that they don't really show your creativity, and named what they are: content farming, which is the thing YouTube doesn't want in the Partner Program.
He also made clear that the policy doesn't care how anything was made. YouTube's policies, in his framing, are independent of the production method — the test is the final content.
Turn the negative around and you get four things the policy rewards.
What the policy rewards
- Variation between uploads
- A narrative arc
- Visible creative decisions
- Value to the viewer
That's not a list YouTube published. It's what falls out of its own description of the problem. The three named categories are covered in our breakdown of the July 2026 update — this post is about what to build once you've read them.
Why does a series satisfy all four?
Variation is structural, not decorative
In a series, each episode has a different scene and a different beat because the story moved. You don't add variation. It's a by-product of plot.
A template channel has to manufacture variation deliberately, which is why it usually doesn't. YouTube's policy page names the failure directly: characters put in the same situation over and over with the same outcome.
An arc is the format's definition
Serialization is what separates a drama from a clip. A viral video is a moment. A short drama carries recurring characters, an ongoing plot, and deliberate continuity from one episode to the next.
That continuity is the thing you can't fake at volume. Episode 40 has to remember episode 3.
Creative decisions are visible in the cut
Where an episode ends. What it withholds. How a beat lands. Every one of those is a choice, and choices are what the policy is asking to see.
Where this falls short. None of it is automatic. A "series" that runs the same room, the same blocking and the same beat 70 times is templated content with a cast list. The format gives you the structure. It doesn't do the work.
Which policy does this actually protect you under?
Worth separating, because they're two different systems with two different bars.
Halprin distinguished them directly. Community Guidelines set the baseline for removal — egregious harm, hate speech, nudity. Partner Program policies apply a higher bar and evaluate the channel as a whole for monetization eligibility.
A scripted series clears the first easily. What it's doing is clearing the second, which is where the inauthentic content policy lives and where channel-level review happens.
He also addressed the thing creators worry about most: flag count doesn't dictate enforcement. Decisions come from comparing the channel's content and behaviours against the policy, not from counting reports.
What does a compliant episode look like?
The format has settled on a set of conventions. They aren't rules, but they're what the audience has trained itself to expect.
- Episode length
- 30–90 seconds
- Script length
- 400–600 words
- Dialogue
- Under 12 words per line
- First 10 seconds
- Re-anchor, don't establish
- Forward moves
- Exactly one per episode
- Ending
- Suspension, not resolution
- Sound
- Captions carry the story
- Performance
- Close and restrained
Three of those deserve explaining.
Dialogue under 12 words a line has three separate reasons behind it. Captions wrap awkwardly past that length. Lip-sync drifts on longer takes. And every extra second of dialogue is a clip you're paying to generate.
The first 10 seconds re-anchor. A returning viewer doesn't need introducing to the premise. They need reminding where they were. Establishing shots are for episode one.
Exactly one forward move. A revelation, a confrontation, or a decision. One. Two makes the episode feel rushed and robs the next one of its reason to exist.
Then end in suspension. Resolution is what makes a viewer stop watching.
no face swaps, no wardrobe changes, no broken hands, no floating props, no camera jumps
Standard negative set for series work. Append it unchanged rather than rewriting it each time.
Where this falls short. These are conventions, not physics. A good reason beats any of them. But break one by accident and it reads as a mistake rather than a choice — which is exactly the distinction the policy is looking at.
How do you keep a cast consistent across 70 episodes?
This is the part that decides whether a series is actually possible, and it's a workflow question rather than a craft one.
Each generation is independent. The model has no memory of the previous shot, so a character re-described in words comes back slightly different every time — and across 70 episodes those small variations compound into a visibly different person.
The mechanism that solves it is a named reference. Create the actor once, save it to your library with an @handle, and call that handle in every prompt instead of re-describing the character. On Hexcoded a saved actor is available across all four tools, so the cast you build for episode one is the cast you call in episode seventy.
The same applies to places and looks. Characters, elements, styles and locations are all reference types, so the apartment your series lives in can be a handle rather than a paragraph you retype and accidentally vary.
Is there actually a US audience for this?
Yes, and it's now the largest market outside China.
The studio behaviour is the more telling signal. Fox Entertainment took equity in Holywater's MyDrama platform rather than running a pilot. ReelShort has partnered with Paramount. Over the same period, traditional streaming app downloads fell.
When studios start placing equity and talent deals instead of experiments, the format has stopped being speculative.
What does this cost you in effort?
The work moves rather than disappearing. It shifts out of the shoot and into pre-production.
Write the series bible
Characters, arcs, tone, world rules. This is the single source of truth for everything that follows.
Create and name the cast
Generate each actor from a description or upload a face, then give it a handle. Do this before you generate footage — the handle is what holds the character together across episodes.
Build the shot list
Break each episode into scenes and shots before you generate anything.
Then generate episodes
Order matters more than speed here. Work out of order and you multiply re-rolls.
The payoff is that episode two starts faster than episode one, because everything you named persists. A mid-series change — a scar, a haircut — means regenerating the reference once rather than editing every prompt that mentions it.
Worth knowing that Plan First generates a storyboard and script preview without consuming credits, so an episode's shape can be checked before anything is spent.
Where this falls short. How much faster depends on your models, your shot count and your tolerance for re-rolls. We haven't measured it, so we're not going to give you a number. Anyone quoting you hours per episode is describing their own tool stack, not yours.
What would make this go wrong?
Three failure modes, and all are avoidable.
A series in name only. Same room, same blocking, same beat, 70 times. A cast list doesn't make it a series. If a viewer can watch three episodes in a row and not tell them apart, you're in category one of the policy.
A drama frame around advice. An AI persona giving financial or health guidance is category three regardless of the fiction wrapped around it. The frame doesn't change the category.
And a distribution mismatch. Worth checking your platform's licence scope early — Hexcoded's, for instance, covers online platforms plus your own sites and apps, with television, print and outdoor needing a separate written agreement. A series built for YouTube or a vertical app is in scope. A broadcast deal isn't, without that conversation.
- Pick the format before you pick the model. Format is now a monetization decision, not just a creative one
- Build a series, not a slate of one-offs. Recurring characters and continuity are what the policy language protects
- The policy is independent of how content is made. AI isn't the test — the final content is
- Move one thing forward per episode. A revelation, a confrontation, or a decision. Exactly one
- End in suspension. Resolution is what makes a viewer stop
- Keep dialogue under 12 words a line. Captions wrap, lip-sync drifts, and clips cost money
- Name your cast and call them by handle. Re-describing a character is what causes drift across 70 episodes
- Never put an AI persona on health or money, even inside a drama. The frame doesn't change the category
- Check your platform's licence scope early. Online is usually covered; broadcast usually isn't
Video that shows variation between uploads, a narrative arc, and visible creative decisions. YouTube's trust and safety lead described the opposite of all three as the target: very similar videos that are generic, lack a narrative arc, and don't show creativity.
The safety is structural rather than categorical. A faceless channel can be perfectly compliant. But a serialized story has to vary each episode because the plot moved, whereas a template channel has to work at variation deliberately, and usually doesn't.
The format has settled on 30 to 90 seconds per episode, with scripts running roughly 400 to 600 words. Series typically stack 70 or more episodes into one continuous arc.
Create each actor once, save it with a handle, and call that handle in every prompt rather than re-describing the character. Each generation is independent, so a re-described character comes back slightly different — and across 70 episodes those variations compound.
Yes, and it's now the largest market outside China. Microdrama app downloads passed 2.3 billion globally in 2025, and the US is forecast to account for around half of microdrama revenue outside China in 2026, reaching $1.5 billion. Fox Entertainment has taken equity in a vertical platform and committed to more than 200 titles.
No. Review is channel-level and other policies still apply, including disclosure and advertiser-friendly guidelines. A series that recycles one scene with a swapped name is templated content whether or not it has a cast list.
Name the cast once, ship 70 episodes
Save each actor with a handle and call it in every prompt, across every tool. Make a Custom Video scripts, casts and renders an episode from a plain description — and Plan First shows you the storyboard before it spends a credit.
Start a seriesMore on workflow, continuity and shot planning in Production.