YouTube’s New AI Editing Tools (2026): What Creators Actually Get
The short version: at Made on YouTube on September 23, YouTube shipped AI feedback on your drafts, three-way dynamic thumbnails, A/B testing for video cuts, and a Gemini conversational editor inside Shorts and the YouTube Create app. Most of it will genuinely save hours. The catch is the other AI YouTube runs: the labelling system that tagged Kurzgesagt’s animation as AI in August is still misfiring, and viewers are now installing extensions to hide anything wearing the label.
Read this before you switch anything on
Use it: Shorts creators, faceless/high-volume channels, anyone with a back catalogue of 50+ videos, teams that can’t hire an editor.
Be careful: narrative and animation channels (the label problem is not solved), and anyone who does not want Google’s AI touching their packaging and their likeness. Comment moderation and old-video thumbnail refreshing are both opt-in — leave them off until you have watched them on one video.
What was actually announced
Most coverage flattened this into “YouTube adds AI.” The version worth reading is the status column: a few of these are live, some are “coming soon,” and two are formally 2027. YouTube’s own post on the day came from Aparna Pappu, one of its VPs, and it is unusually specific about the sequencing.
| Feature | What it does | Status |
|---|---|---|
| Ask Studio | Answers questions about your channel, summarises comments, and gives feedback on a draft — pacing, structure, storytelling. Also lands in the mobile app. | Rolling out, Studio web + mobile |
| Channel-matched thumbnails | Generates thumbnails in your channel’s existing visual style. | Live |
| Dynamic thumbnails | Upload up to three thumbnails per video; YouTube picks which audience segment sees which. | Live |
| Video A/B testing | Test up to three cuts of one video to find the hook that holds viewers. | “Coming soon” per YouTube; the Shorts version in 2027 |
| Conversational editing | Gemini-powered editing in Shorts and YouTube Create: reorder frames, trim, sync music to the beat, switch back to manual at any point. | Announced; Gemini was already in Shorts/Create earlier in 2026 |
| Catalogue refresh | Ask Studio can run in the background and propose or test new thumbnails on older videos. | Opt-in, later this year |
| AI comment moderation | Learns your moderation style and applies it for you. | Testing, opt-in |
| Voice + likeness detection | Speaking-voice detection gets folded into facial detection; likeness tool moves into the mobile app. | Later this year; deepfake detection already open to eligible channels |
| Shopping affiliate | Program expands to 35 countries by year-end; AI auto-tags products; Amazon tags open up in India and Brazil. | Rolling out through 2026 |
Sources: YouTube’s own announcement (2026-09-23), TechCrunch, Quartz, Reuters-reported shopping figures. Live auto-dubbing is the other 2027 item worth knowing about if you publish in more than one language.
The numbers YouTube chose to put in its own post
Three of them matter more than the feature list:
40 million A/B experiments run by creators on titles and thumbnails since the feature formally launched in 2024, according to YouTube. That is the reason dynamic thumbnails exist at all — the company already knows testing wins.
72% of US video posters aged 14–44 had used AI to help create or edit in the past year, from YouTube’s first GenAI trends report. This is the sentence that tells you who the new Studio tools are built for. It is not a marketing claim, it is a census of their own uploaders.
Hundreds of thousands of channels per day used Gemini inside Shorts and YouTube Create in August 2026 alone, per YouTube. Whatever you think of AI editing, it is no longer an experiment on that platform.
The other AI YouTube runs — and it is not going well
On August 7, Kurzgesagt — a human animation studio with a staff of illustrators — posted that YouTube’s AI detection had flagged them, under the headline that the detection “kicked us in the face.” In September, a Safari extension called Weedout that simply hides AI-labelled videos hit 185 points on Hacker News. The comments there are the most useful thing written about YouTube’s AI labels this year, because they are from people who want the label to work.
“My only concern is that I have seen legit videos by creators on the platform for many years, with the tiny AI (i) label below the video. I have seen this even on channels that are not faceless. You would think if anyone could get this right, it would be Google/YouTube, but apparently not.”
“I’ve found that the AI label doesn’t necessarily mean AI slop, it just means there’s a non zero amount of AI in the video. I want to avoid AI slop but not all videos that have an AI title screen or whatever.”
The detection itself is the weak link, not the concept. As one commenter with experience across platforms put it, detectors are “surprisingly bad” — non-AI animated and cartoon content gets mislabelled more often than obviously synthetic footage, because the classifiers lean on metadata and surface style rather than provenance (dathinab, HN).
And the audience has already started filtering. The most common reason people gave for installing a hider was not slop — it was synthetic voices: “once I detect that the voiceover is AI generated that shit get instablocked forever” (karim79, HN). Someone else made the sharper point about who this costs: “if you’re technically competent, you just write yourself an extension to block the slop spam... so in the end it’s rotting much more the brains of the least prepared” (negura, HN).
So what do you do with it on Monday
- Turn Ask Studio’s draft feedback on for one video. Test it against your own judgement once. Its advice is built from your channel’s history and comments — it will be confidently wrong about a format change you are deliberately making.
- Dynamic thumbnails are the free win. Three existing designs, no change to the video, YouTube does the split testing. There is no content-integrity question here, which is why it is the safest thing in the list.
- Leave the catalogue refresh off until you have watched it. Letting an assistant rewrite thumbnails on videos that already earn is the highest-risk item on the page, and it is opt-in for that reason.
- If you use an AI voice, assume the label, and price the trade. You are buying reach with a filtered-out segment attached. That is a business decision, not a moral one — just make it on purpose. Our look at voice-cloning alternatives covers the cost side of that decision.
- Check your own channel for the label. Animation, documentary, and heavily-stylised channels are the ones with a documented misfire problem. If you find one, you are not imagining it — Kurzgesagt had the same fight with an audience 40 times your size and still had to escalate publicly.
The honest verdict
If you publish shorts or run a faceless channel: use it.
Draft feedback, dynamic thumbnails and three-cut testing are aimed squarely at your workflow, and the 72% number says your competitors already moved.
If you are a writer, animator or documentary maker: use the testing tools, avoid the packaging generators.
You will not be labelled for using dynamic thumbnails. You may be labelled for using AI illustrations or voice, and the label currently has an accuracy problem that YouTube has not fully acknowledged. That is the whole risk calculation, and it is unusually easy to get right: test the packaging, keep the craft.
And the uncomfortable part.
The same company that just handed you a creative partner already mislabelled a studio that helped define the platform. Trust the A/B tests. Do not yet trust the classifier.
Voice cloning alternatives → Screen recording, reviewed →
FAQ
Are YouTube’s new AI tools free? YouTube has not put a separate price on any of them — they are Studio and YouTube Create features, not a new subscription. The monetisation is elsewhere in the same announcement: the shopping affiliate programme expanding to 35 countries by the end of 2026, with more than 1.3 million creators already enrolled.
When does video A/B testing arrive? YouTube says “coming soon” for testing up to three cuts of a video; the Shorts version was given a 2027 date in TechCrunch’s coverage of the event. Title and thumbnail A/B testing has been live since 2024.
How do I know whether my video has an AI label? We could not find a documented per-video label dashboard while putting this together, so treat that as a gap: creators in the threads above noticed it by looking at their own video pages. What is documented is that channels are eligible to sign up for deepfake detection and can manage their likeness in Studio, and that voice detection is being folded into it later this year.
Will the AI editor replace my editor? Not on the evidence in YouTube’s own post: the conversational editor handles trimming, frame reordering and beat-syncing, and manual editing stays available throughout. It is a first-draft tool, and it is priced like a feature, not a hire.
Method: YouTube’s Made on YouTube 2026 announcement post (2026-09-23, Aparna Pappu), plus TechCrunch, Quartz and Tubefilter coverage of the same event, checked 2026-09-24. Community quotes are verbatim from two Hacker News threads — the Weedout extension discussion (2026-09-01, 185 points) and the Kurzgesagt detection thread (2026-08-07) — with comment links. We do not take payments from vendors.