YouTube's AI Can Skip the Video for You. Creators Get Paid When You Don't.
YouTube is testing a feature that lets you ask it a question in plain English and get a text summary back - with cited videos, timestamps, and follow-up questions - without ever clicking into a video. At the same time, it already has a button inside most videos that lets you ask Gemini what the video contains and get timestamped steps without watching the whole thing. Two features, same direction: you get the answer faster, the video gets fewer views.
Both of these are real, both are live in some form right now, and both are being discussed almost entirely from the viewer's perspective. The creator side of this is quieter. It shouldn't be.
Two features, one direction
The first feature is the Ask button - sometimes labeled "Ask Gemini" - that appears inside the YouTube player. You're watching a tutorial on how to set up a home network. You click the button. Gemini opens a chat panel that already knows the video's content. You type "give me the steps in this video" and get back a bulleted list with timestamps you can click. You land at the right moment. You don't have to sit through the 4-minute preamble about the creator's setup or the sponsor read.
Sydney Butler at HowToGeek described it as "probably one of the best features for users YouTube has added in ages" - specifically for tutorial content, where most videos are padded with irrelevant framing and you just want the procedure.
The second feature is called Ask YouTube. It replaces the search bar on YouTube.com with a conversational interface. You type a question instead of keywords. YouTube returns an AI-generated text summary, a primary cited video linked directly to a timestamped section, and galleries of related long-form videos and Shorts. You can ask follow-ups in the same thread. You get comparison tables. You get structured answers pulled from multiple videos.
Search Engine Journal's Matt Southern tested it in April 2026. He searched for reactions to a specific AI model release, got a generated title, a subhead, a summary paragraph, and an embedded video timestamped to a relevant section - plus citations and related Shorts. He then asked a follow-up: "how does it compare to [another model]?" and got a comparison table with links to the source videos. All from the search results page. None of it required watching a video.
YouTube describes Ask YouTube on its Premium Early Access page as "a new way to search on YouTube that feels more like a conversation." The experiment is available to US Premium subscribers aged 18 and older, searching in English on desktop, through June 8, 2026.
The business model contradiction
YouTube's revenue runs on ads. Ads run during video playback. No playback, no ad impression. It is not a complicated equation.
The Ask button reduces playback on individual videos - specifically the parts creators were paid to make. The rambling intro that gets skipped via AI timestamps was also the part where the pre-roll ad ran. The sponsor segment that gets jumped with a single Gemini query was the part the creator actually negotiated money for.
Ask YouTube goes further. If you search, get an AI summary, read the answer, and leave - YouTube served you content from multiple creator videos without triggering a single view. No views, no watch time, no revenue split. The creator whose video was cited as the "primary source" got zero.
YouTube spent a decade training creators to think watch time was the most important metric. Now it is building features that bypass watch time entirely.
YouTube's own A/B title testing - which I wrote about in April - optimizes for "watch time per impression," not just click-through rate. YouTube's logic was: a title that gets clicks but doesn't hold viewers isn't actually a good title. The metric it picked was watch time. The features it is now shipping reduce watch time. That's not a coincidence. It is a pivot.
Ad impressions and watch time credit only accrue when video content is playing. AI summaries drawn from video content generate no revenue for the creator whose content was used.
Source: YouTube monetization model; revenue signal proportions are illustrative
So why is YouTube building this?
That is the right question. YouTube is not naive about its own business. The people who built Ask YouTube understand ads. They understand CPMs. They understand that fewer views is not a goal.
The answer is that YouTube is not comparing "AI summary" to "full video watch." It is comparing "AI summary on YouTube" to "AI summary somewhere else." ChatGPT can already summarize YouTube videos. Perplexity indexes YouTube content. Gemini.google.com can pull from YouTube transcripts. If someone goes to a competitor to process YouTube content, YouTube gets nothing - no session, no data, no ad impression, no subscription renewal signal.
Keeping the AI interaction inside YouTube.com means YouTube retains the session. It serves the summary. It controls what gets cited. It can potentially show an ad before or after the AI response. It keeps you logged in. That is worth something, even if it is worth less than a full video view.
This is the same calculation YouTube made with the Shorts time limit - build the user control feature yourself before regulators force a worse version on you. Build the AI search feature yourself before a competitor takes the session entirely.
What creators should actually do with this
Complaining that AI summaries reduce watch time is accurate but not actionable. The features are being tested. They are likely to roll out more broadly. The question is how to make content that works in both environments.
The Ask button rewards dense, well-organized content. A video with clear chapter timestamps, a tight structure, and a procedure you can actually extract will generate better AI summaries. Better summaries make people want to see the whole thing, not replace it. The curiosity gap that works in thumbnail design also works in AI summaries: if the summary is interesting but feels incomplete, the viewer clicks through.
The Ask YouTube citation question is harder. YouTube has not shared what makes a video the primary citation versus a supporting one. What we know from every other YouTube ranking signal is that engagement depth matters - watch time, likes, comments, shares. A video that holds people when they do click in is the kind of video that gets surfaced. The behavior that earns citation is the same behavior that earns everything else on this platform.
The deeper problem is that "watch time" was always a proxy. It was a proxy for value - the assumption that if someone kept watching, the content was worth watching. AI summaries expose that proxy for what it is. A viewer who gets their answer from a 30-second Gemini response and a viewer who sat through a 12-minute tutorial and retained nothing generated the same watch time signal: zero and twelve minutes respectively. Only one of them got value.
The features YouTube is building measure engagement with AI responses - thumbs up, thumbs down, follow-up questions. If those engagement signals eventually influence which creator content gets surfaced, the creators who actually answer questions clearly will do well. The ones who optimized for watch time by padding intros and burying the answer will not.
Watch time was the scoreboard. It was always showing the wrong number. The AI features just made that more visible.
Ask YouTube is currently an experiment available to US Premium subscribers through June 8, 2026. The Ask (Gemini) button is available in select regions and languages in the browser player and YouTube app. Neither feature has a confirmed creator monetization model for AI-cited content yet.
Join the conversation