Should I reveal the parts touched by AI on YouTube?
YouTube's disclosure obligation is not whether it used AI, but whether it made something that wasn't real look realistic. Subtitle creation, color correction, upscaling, and audio restoration are not public domain, while AI-generated music is public domain. We looked at the broadcast editing work one by one, and even summarized the sanctions for not disclosing and the differences in monetization review.
ClipSpray Team
Key takeaways
- The standard for disclosure is not whether AI is used, but whether something that is not real is made to look realistic. The nature of the result, not the tool, is what is judged.
- Items that YouTube has specified as exceptions include subtitle creation, color correction and lighting filters, special effects filters, video sharpening, upscaling and restoration, and voice and audio restoration. This is what most automatic editing tools do.
- Conversely, AI-generated music is explicitly subject to public disclosure. This is the item that editors encounter most often, so it is safer to set it as a rule.
- Failure to disclose may result in direct labeling, content removal, or suspension from participation in the YouTube Partner Program.
- Disclosure obligations and monetization review are separate. Disclosure is a matter of displaying a single video, and decisions about mass production or reuse are made after looking at the entire channel.
When you start using automatic editing tools, you have one more thing to worry about. Should I indicate that this was made with AI? YouTube's standard is not whether it uses AI, but whether it makes something that is not real look realistic. In this article, we will use official examples to see where the line is drawn and tackle broadcast editing tasks one by one.
What does the disclosure obligation look like?
Look at the nature of the result, not the tool.

Altered or composite content disclosure documentAn example of what was disclosed is as follows.
- “Generating additional videos of real places with AI” one case
- “Making it look like someone gave advice they didn’t actually give.”
- “A realistic depiction of a tornado or other weather phenomenon approaching a real city that did not actually occur.”
- “AI generated music”
I can see what the four items have in common. This is when something is made to look like it happened when it didn't actually happen. There is nothing here to say that editing was done quickly or automatically.
What do I need not disclose?
The same document lists the exceptions directly, and that list largely overlaps with the edits.
Items listed as items that do not need to be disclosed “Apply cosmetic filters”, “Color correction or lighting filters”, “Special effects filters (e.g. add background blur or vintage effects)”, “Create subtitles”, “Sharpen, upscale, restore video, or restore speech/audio” It is.
It is particularly important that subtitling is listed on the exception list. Creating subtitles using voice recognition is a major part of automated editing, which means there is no need to mark it as public. How to improve subtitle quality Automatically generate shorts subtitlesI have organized it separately.
What happens if you tackle broadcast editing tasks one by one?
Most of them are not public, and there are only two points of divergence.

| What we do in editing | public display | basis |
|---|---|---|
| Select highlight sections from long replays | No need | This is not a task of creating a scene that did not exist before. |
| Cut editing and reordering | No need | Layout of actual scenes |
| Create voice recognition subtitles | No need | Subtitle production is on the exception list |
| Vertical conversion, image quality improvement, sound cleanup | No need | Upscaling and audio restoration are on the exception list |
| Installing AI-generated background music | needed | Examples of AI-generated music being made public |
| Synthesize voices to create words that never existed before | Necessary and dangerous | Make it look like you said something you didn't say |
The last line is not a matter that will end up on public display. If you make up words that you did not say, even if it is your own voice, it becomes content that misleads viewers, and if it is someone else's voice, it also raises issues of portrait and voice rights. This is an item that started out as fun but then quickly became the biggest issue.
Why is AI-generated music taken separately?
This is because it is directly written in the example for disclosure.
In editorial practice, this is the most frequently encountered item. These days, tools that turn BGUM into AI are common, and they are easy to use because you don't have to worry about royalties. But Royalty issues and disclosure obligations are different issues. Even if the copyright is clear, music created with AI is subject to disclosure.
When it comes to shorts, there is one more judgment involved. When you put music in, half of the revenue associated with eligible views goes to licensing fees, reducing creator revenue from 45% to 22.5%. In other words, the choice to include AI music affects both public display and revenue distribution. Which music to choose Choosing Shorts MusicThis is covered in detail.
If I make it public, where will it be displayed, and what happens if I don't?
Labels are placed where viewers see them, and failure to display them will result in sanctions.
When published, the document notifies "A label will be added to the video", followed by "The label may appear in the video player", "The label may appear in the expanded description box".
This is the case when it is not displayed. “You may be subject to direct labeling, removal of your content from YouTube, or suspension from the YouTube Partner Program.”
Of note is the first item. Even if you don't mark it, YouTube may label it for you. In other words, the actual benefit of choosing to hide it is not great, and instead, there is only the risk that it may lead to suspension of the partner program.
What does this have to do with monetization review?
It's separate. The viewing and judgment units are different.

| Category | Disclosure of altered or synthesized content | Channel Monetization Policy |
|---|---|---|
| seeing | Did you make something that wasn't real look realistic? | Have I added anything to the original, and are the images distinct from each other? |
| Judgment unit | One video | All Channels |
| response | Show when uploading | Change the editing and composition itself |
| When it's misaligned | Apply labels, delete, stop YPP | Monetization application rejected |
The same thing is that both do not take issue with the tool. The example of the channel monetization policy was “AI-generated content that gives the impression of being mass-produced and created without adding the creator’s original and authentic insight or perspective,” not the fact that AI was used per se. These judgment criteria are summarized in AI-edited broadcast clips, , the line to be judged for monetization.
What do you check before publishing?
Checking twice is enough. Once when creating and once when uploading.
1️⃣ Check to see if you created something that didn't exist before: See if this video contains scenes that didn't actually happen or words that weren't said. If not, it mostly ends here.
2️⃣ Check the source of the music: Music generated by AI is subject to public disclosure. This does not apply to audio library music or original broadcast sounds.
3️⃣ Displayed on the upload screen: If applicable, check the disclosure item in the upload process. It's easier to do it at the beginning than to add it later.
4️⃣ Take a look at each channel as well: Whether the ten recent videos are distinct from each other is a matter of monetization, separate from the obligation to disclose.
5️⃣ Fix it as a rule: Do not use voice synthesis, and if you use AI music, make sure to display it as a channel rule so you do not have to judge it every time.
The job of automation is at the front end, where candidates are found and refined, and it is up to humans to decide what to make. The reason ClipSpray leaves a timeline report showing what signal was selected for each candidate section is so that a person can make that judgment.
Frequently Asked Questions
If I create a clip with an AI editing tool, do I have to mark it as public?
Most of the time you don't need it. Examples of what YouTube has targeted for disclosure include AI-generating additional footage of real locations, making it look like someone said something they didn't say, or realistically depicting an event that didn't happen. Selecting sections, cutting them, and adding subtitles are not included here.
Are automatic subtitles also public?
No. Subtitle production is included in the example that YouTube directly specifies as an item that does not have to be disclosed.
What about improving the picture quality or refining the sound?
It is not subject to public disclosure. Video sharpening, upscaling, restoration, and voice or audio restoration are also listed as exceptions.
What happens if I add background music created by AI?
It must be disclosed. AI-generated music is listed in the public disclosure examples. In Shorts, adding music itself affects profit distribution, so it is better to judge the two together.
When I make it public, where will it appear?
A label is added to the video and may appear in the video player or expanded description. It is attached in a location where viewers can see it.
Is public display detrimental to monetization?
Public display itself is not indicated as a reason for demonetization rejection. Rather, if it is not disclosed, disadvantages such as direct application of labels, deletion of content, and suspension of participation in partner programs are clearly stated.
Source
Don't miss the moment when my broadcast explodes.
Just enter a replay link and the chat and sponsorship responses will be analyzed to create scenes to be made into shorts. I'll find it for you.
Try it for free with my stream10 vouchers upon signing up · No card registration
Share this post
read together
Creating a reel from a 3-hour broadcast: How to choose 15 seconds to write and move them vertically
2026-08-31 · About 12 min
How to find 30 seconds in a 6-hour broadcast and reduce streamer reel editing time
2026-08-31 · About 10 min
Reasons why editor recruitment is inconsistent, from division of work to unit price and settlement
2026-08-31 · About 12 min


