YouTube Generative AI Strategy: Opportunity Map, First Bet and Guardrails

You are a product manager at YouTube. Generative AI is changing how video is created, found and watched. What should YouTube do with it? Recommend a strategy: where YouTube should apply generative AI, which opportunity it should pursue first, and how you would know whether it is working. ```hint Start from YouTube's sides YouTube serves viewers, creators and advertisers, and generative AI both helps and threatens each of them. Pick an objective, compare opportunities across those sides and their risks to trust, and only then choose a first bet. ``` ### Clarifying Questions - Is the focus on user-facing products for viewers, creators and advertisers, or should internal uses such as moderation and support count as well? - Which goal matters most here: creator growth, viewer engagement, ad revenue, or trust and safety? - Is this a set of near-term launches or a multi-year strategy? - Are there limits on using creators' videos to build or tune models, given creator and rights-holder concerns? ### What a Strong Answer Covers - A stated objective and an inventory of opportunities across viewers, creators, advertisers and platform integrity - A prioritization method applied to YouTube-specific opportunities, ending in one clear first bet - Enough detail on the first bet to see who it serves, how it works and what ships first - Risks particular to generative AI on a video platform: synthetic or misleading content, creator consent, floods of low-effort uploads, and inference cost - Success metrics with guardrails ### Follow-up Questions - AI-generated uploads flood the platform and viewers complain about low-quality videos. What do you change? - A prominent creator objects to their videos being used to tune the models behind your feature. How do you respond? - Generating content costs far more to serve than recommending it. How do you decide which generative features are worth their compute cost?