Was your YouTube channel mistaken for AI content?
Start by separating a visible AI label, a formal monetization decision, and a reach drop with no notice. Then use the matching correction, appeal, or evidence-building route. The historical case below separates what Google, YouTube, and Kurzgesagt actually said from what creators can only infer.
Start with the notice—not a suspected cause.
A visible AI label is not a YPP suspension, and a reach drop alone proves neither. YouTube provides routes for correcting eligible automatic labels and appealing formal YPP decisions. It does not document a self-serve check or appeal for an undisclosed distribution restriction. Choose your route from the notice you actually received, not the cause you suspect.
START HERE · THREE DIFFERENT ROUTES
What did YouTube actually show you?
A visible AI disclosure label
Review the video's notice and disclosure. An eligible automatic label can be corrected if it is wrong; that is not a monetization appeal or proof of a reach penalty.
Check the label correction routeA YPP suspension or application rejection
Read the reason, deadline, and available appeal route in Studio's Earn section. Correcting an AI label is a different process.
Use the formal YPP appeal routeLower exposure, with no relevant notice
This is an analytics investigation, not a confirmed AI classification. Compare similar uploads and traffic sources before naming a cause or sending a support case.
Check the evidence firstStudio's current help calls the column Notices: open the Content tab, inspect a notice's Impact summary, and use Action or Review where offered. The August timeline below preserves the earlier “Other notices” wording. These are not three names for the same restriction. YouTube Help: Content tab notices and review actionsYouTube Help: AI disclosure and automatic labels
WHEN YOU HAVE A FORMAL NOTICE
Have a YPP notice? Use the formal appeal route
- Open Earn in YouTube Studio. Save the notice and deadline; follow Start Appeal for the offered video or Creator Support route.
- For an invited pre-suspension appeal, YouTube gives a 7-day window. If you did not appeal before suspension, you have 21 days after suspension. For an application rejection, follow your own notice.
- Do not delete videos before submitting an appeal: YouTube reviews the channel in its current state. If a video appeal is offered, demonstrate the channel's production process against the cited policy.
YouTube Help: appeal a YPP suspension or application rejection
A label correction is not a substitute for this appeal. YouTube's channel monetization policies separately address generic or repetitive content and reused content. Use the specific reason in the decision; do not reduce every rejection to “AI is banned.” YouTube Help: channel monetization policies
WHAT REALLY GETS FLAGGED
What actually puts monetization at risk
This is a YouTube Partner Program eligibility rule, not an AI disclosure label or proof of a hidden reach penalty. YouTube's July 15, 2025 update log says it renamed “repetitious content” to “inauthentic content” to clarify that repetitive or mass-produced content was already ineligible for monetization. The official source does not describe a new August 2026 rule or effective date. YouTube Help: channel monetization policies
| Creator shorthand | YouTube's actual rule | What that means |
|---|---|---|
| “Image slideshows with no human narration” | YouTube lists image slideshows, templated storylines, or scrolling text with minimal or no narrative, commentary, or educational value. YouTube Help: channel monetization policies | Being faceless or using a slideshow is not the standalone test. The risk is the missing value and interchangeable template. |
| “The same character in the same situation” | The policy names characters put in the same situation repeatedly with the same outcome through a highly similar storyline template. It separately allows a recurring cast when each video has a distinct storyline, focus, or concept. YouTube Help: channel monetization policies | Reusing a character is not enough. Repeating the same story and outcome with little meaningful variation is the risk. |
| “AI-generated videos” | YouTube names AI-generated content made with generic or unoriginal templates that appears mass-produced and adds no original, authentic insight or perspective. YouTube Help: channel monetization policies | AI use alone is not the rule. The policy combines a generic template, a mass-production pattern, and a lack of original creator value. |
01 · VERIFIED TIMELINE
What happened—and what has not been confirmed
The most important correction is that two separate disclosures have been blended into one story. Google described a system for terminating coordinated synthetic-spam clusters. Kurzgesagt separately said YouTube confirmed that automatic AI detection had misread its human-made videos. No public source says the cluster system was the system that affected Kurzgesagt.
Google describes cluster-level synthetic-spam enforcement
Google Research lists only the year “2026”; it does not provide an exact official publication day. The Google-hosted PDF embeds a May 14 CreationDate, but file metadata is not a publication date. An archived version of the research catalog says the system terminated 50,000 clusters comprising 130,000 channels over a six-month period. The current downloadable PDF explains the architecture but does not contain those two counts.
Google-hosted S-CTS paper (PDF)Archived Google Research catalog entry
YouTube announces expanded automatic AI labeling
TeamYouTube said it was expanding automatic detection and placement of AI disclosure labels. It also said a label by itself would not affect recommendations or monetization. This announcement is about visible labels; it does not describe the cluster-termination system or a hidden reach penalty.
Kurzgesagt publishes its first-hand account
Kurzgesagt said its superpredator upload became its worst performer since 2013 even though people clicked more often, watched longer, and responded positively. The studio said its YouTube contacts confirmed that automatic AI detection was wrongly treating its human-made videos as AI slop.
Kurzgesagt's original statement on XKurzgesagt's YouTube community post
The only public fix update is explicitly tentative
Kurzgesagt did not publish a separate, definitive repair notice. In the original July 31 statement, it said it
hopefully
had been fixed for the moment. As of this guide's August 20 check, we found no later first-party post confirming a permanent fix and no incident-specific statement from YouTube.YouTube explains how creators can correct most automatic labels
TeamYouTube added instructions for finding automatically labeled videos in Studio's “Other notices” column and, in most cases, changing the AI-use survey to “No.” This is the closest official correction flow, but YouTube did not connect the update to Kurzgesagt or say that correcting a label reverses a distribution problem.
TeamYouTube's AI-label announcement and August updateYouTube Help: AI disclosure and automatic labels
02 · HOW THE SYSTEM WORKS
It looks for an operation, not just one suspicious video
The Google paper calls its design the Scalable Cluster Termination System, or S-CTS. It combines account relatedness with a separate synthetic-content classifier. Upload pace is one signal among many, not a standalone rule.
Link related accounts
A bot-network classifier groups accounts using proprietary infrastructure and behavior signals resembling shared API usage, event timing, generation metadata, and inorganic activity patterns.
Measure synthetic-pattern prevalence
A content classifier reviews multimodal context: frames, audio, transcripts, title and description terms, visual embeddings, templated narratives, and channel upload pacing.
Act where both signals intersect
The system targets high-confidence coordinated clusters that also contain enough synthetic-pattern content. High-confidence cases can be automated; ambiguous ones are routed to human review.
The paper says the cluster requirement is intended to protect isolated creators who experiment with AI. It does not publish a creator-facing list of signals, thresholds, cluster membership, or a way to check whether a channel was scored. Google-hosted S-CTS paper (PDF) Separately, the archived catalog entry describes deployment on a “major Online Video Platform” without naming YouTube. Archived Google Research catalog entry
03 · WHAT TO LOOK FOR
The tell is a mismatch between response and reach—not low views alone
Kurzgesagt's evidence was internally contradictory: the people who received the video clicked and watched, while the video received far less distribution than that response would normally suggest. For your channel, look for several signals lining up across comparable uploads.
| What you see | What it tells you | Next check |
|---|---|---|
| Impressions fall sharply; CTR and average view duration hold or improve | A distribution anomaly is worth investigating, but this is not proof of an AI flag. | Compare Browse and Suggested impressions at the same 24-hour, 48-hour, and 7-day age. |
| Several recent uploads show unusual reach fluctuations | A channel-level issue becomes more plausible than one weak topic. | Compare only videos with similar format, length, topic demand, and release timing. |
| Studio shows an automatically applied AI label | You have a visible classification event, not evidence of a reach penalty. | Inspect the video's notice, review AI use under Attributes, and correct an eligible automatic label if the disclosure is wrong. |
| CTR, retention, and returning-viewer response also fall | An ordinary topic, packaging, or satisfaction problem is more likely. | Fix the title, thumbnail, opening, or audience fit before escalating a platform bug. |
04 · WHAT YOU CAN DO
Build a support case that another person can reproduce
Small creators may not have Kurzgesagt's direct contacts, but they can reduce guesswork. The goal is not to prove the hidden cause yourself. It is to preserve a clean before-and-after record and make one precise escalation easy to investigate.
- 01
Freeze the evidence before changing anything
Export or screenshot the video URL, publish timestamp and timezone, impressions, CTR, average view duration, retention, traffic-source mix, and any Studio notices. Capture the same windows for five to ten genuinely comparable uploads.
- 02
Separate exposure from audience response
Write one sentence that quantifies the mismatch—for example, impressions versus your comparable-video median, alongside CTR and view duration. Avoid “my channel is shadowbanned”; it names a cause your data cannot see.
- 03
Check visible AI labels and disclosures
In Studio's Content tab, inspect the video's Notices and Impact summary. If an automatic AI label is wrong, YouTube says it can usually be corrected by selecting “No” in the AI disclosure survey under Attributes (AI use). Labels from YouTube's own AI tools, C2PA metadata, or manual review cannot be adjusted this way. Disclose content that meets the requirements; do not change an accurate answer just to remove a label.
YouTube Help: AI disclosure and automatic labelsYouTube Help: Content tab notices and review actions
- 04
Preserve proof of how the video was made
Keep scripts, dated drafts, project files, raw recordings, edit histories, asset licenses, and production invoices. This material does not guarantee a reversal, but it gives a reviewer something more useful than a denial.
- 05
Escalate through every route you actually have
Eligible creators can use Creator Support in YouTube Studio. Other creators can send a feedback report with screenshots and contact @TeamYouTube publicly with the channel or video URL—never with private account information. YouTube says feedback may not receive a reply, so keep the submission date and any case ID.
YouTube Help: contact Creator SupportYouTube Help: contact @TeamYouTubeYouTube Help: send a diagnostic feedback report
- 06
Do not delete and re-upload before you preserve the case
Deleting can remove the cleanest evidence and resets the upload. Kurzgesagt took its video offline after reaching YouTube and while planning a revised upload; that is not a general recovery instruction. Export first, file the case, then decide whether a re-upload serves viewers.
For a reach investigation without a formal YPP decision. This note does not replace Start Appeal or its deadline.
Subject: Possible erroneous automatic AI classification affecting distribution Channel / video URL: Published at (include timezone): Change first observed: Comparable-video baseline: Actual impressions at 24h / 48h / 7d: CTR and average view duration versus baseline: Traffic sources that changed: AI disclosure / Studio Notices status: Human-made production evidence available: Request: Please check whether an automatic synthetic-content or integrity classification is limiting distribution, and confirm the appropriate review path.
05 · SOURCE LEDGER
Primary and first-party sources
The incident-specific claim comes from Kurzgesagt, not a public YouTube statement. YouTube's sources below document labels and support routes, including formal YPP appeals. Google's sources document the separate cluster system. Label, Notices, and YPP instructions were checked on September 10, 2026. The monetization examples above were checked on September 16, 2026; the incident timeline retains its August 20 evidence cutoff.
- Google-hosted S-CTS paper (PDF)
- Archived Google Research catalog entry
- Kurzgesagt's original statement on X
- Kurzgesagt's YouTube community post
- TeamYouTube's AI-label announcement and August update
- YouTube Help: AI disclosure and automatic labels
- YouTube Help: Content tab notices and review actions
- YouTube Help: appeal a YPP suspension or application rejection
- YouTube Help: channel monetization policies
- YouTube Help: contact Creator Support
- YouTube Help: contact @TeamYouTube
- YouTube Help: send a diagnostic feedback report
Audit what is visible. Verify the private metrics in Studio.
Kabo's YouTube Channel Audit snapshots public channel and recent video evidence. It cannot see CTR, watch time, traffic sources, or a hidden platform classification—and it will never pretend that it can. Use it as the public half of your evidence packet.