YouTube's Made on YouTube event pushed AI into drafting, thumbnails, testing, and editing. Elsewhere, likeness detection is coming to mobile, with voice detection due later in 2026.
YouTube held its annual "Made on YouTube" event late September, and for creators the main thing it did was insert AI into pretty much everything. Following a year where it is fair to say it has made monetization on the platform more difficult by raising the barriers to entry for the Partner Program (starting next year) and fiddling with view counts, it’s now giving creators a stack of AI tools to make hitting those targets theoretically a bit easier.
A lot of this will be familiar to anyone who has used the latest generation of AI editing assistants, but it is still surprising to see them embedded so tightly into the YouTube workflow. A new draft feedback tool analyzes unpublished videos and offers notes on pacing, structure, and storytelling before a creator hits publish. Thumbnail and title generation now produces options matched to a channel's existing style, and A/B testing is expanding to cover opening hooks as well, with up to three variants tested per video.
There is obviously appetite for all this. YouTube says that more than 40 million A/B tests have been run on titles and thumbnails to date, and later in 2026 it plans to go further still. Next is a system that monitors thumbnail performance automatically and swaps in a better one if the current choice underperforms, without a human being involved at any point in the process.
Perhaps more likely to have an impact is another foray into generative AI territory and a new conversational editing feature rolling out across Shorts and the YouTube Create app. Gemini Omni has been in both since May, and now YouTube is doubling down on its use. Instead of working with a manual timeline, creators can describe what they want in chat and the tool trims pauses and reorders clips for pacing. We have no idea how well it works as yet, but given the potential resources YouTube can throw at something like this, and the potential volume of people who can use it, we expect progress to be rapid.
As well as embracing AI to a surprising degree as a production tool, YouTube used the event to announce new guardrails against its misuse.
Likeness detection launched to Partner Program members in late 2025, was extended to politicians, officials, and journalists in March 2026, and is now open to all adults aged 18 and older. It targets deepfakes and helps creators find content on YouTube where their face appears to be altered or generated by AI. YouTube says it is coming to mobile phones, so creators can enroll and act on match alerts on the go. Speaking voice detection will also be integrated later this year to improve match accuracy. If any relevant infringing content is found, creators can review and decide what to do, including requesting removal through the usual privacy complaint process.
It's a direct response to the deepfake and voice-cloning problem that has dogged the platform for a while now. In many ways it is the logical extension of July's clarification of the company’s inauthentic content policy. That spelled out three specific categories that lose monetization: generic, templated content (including AI-generated video that adds no original perspective), emotionally manipulative or shock content, and AI personas posing as human experts on health, legal, financial, or political subjects.
YouTube goes as far as naming AI "doctors" and AI "lawyers" as examples, which suggests that it’s trying to get ahead of a wave of toxic AI slop before it starts to swamp the platform.