Popular science YouTuber Hank Green apologized to fans for relying too heavily on AI. Although Green stressed that he writes his own scripts, he reconsidered his process after viewers complained about what they perceived as AI influence in his work. In a Reddit post published on July 31, Green said he had been using AI extensively as a research aid. It helped him find academic papers and other resources quickly, including work he had not known existed, but he concluded that the efficiency had harmed his work by leaving him less freedom to discover his own paths into and around a subject. The pressure to keep producing contributed to the problem. Green used AI to locate papers and learning resources until he was moving so quickly that his own process was no longer clear to him. He ultimately concluded that making more things did not make him produce better things. Green also said the dopamine he received from interacting with large language models and continually doing more was neither healthy for him nor good for the world. His concern was not simply whether AI produced correct information, but how early reliance on it could shape research, ideas, and the creative process itself. That concern exposes a gap in YouTube’s AI disclosure policy. The platform requires disclosure when realistic synthetic content makes a real person appear to say or do something they did not, or depicts a realistic scene that never occurred. It does not necessarily require disclosure when AI assists with idea generation, outlines, scripts, thumbnails, titles, or infographics. The policy also permits creators to clone their own voices for voiceovers without mandatory disclosure. An AI-generated missile animation in a fully animated video likewise needs no label, while AI-generated music does. These boundaries focus on realistic deception rather than the extent to which AI shaped a project. The gap becomes more consequential in the example of a 30-minute video intended to influence opinions about geopolitics. AI could generate its premise, conduct research, prepare the outline, and help draft a script read by the creator’s cloned voice, yet that extensive involvement might not require disclosure. A human researcher could reach sources by different routes, process the material more deeply, or notice unexpected connections, while an AI-generated outline could place the work on a predetermined track too early. Somali creators who use AI to speed up research and video production face the same trade-off: the tools can reduce workload, but excessive reliance can narrow the distinctive voice and perspective behind a story. Because YouTube’s labels do not capture every form of AI assistance, creators still have to judge how much of their work’s direction they are handing to the machine.