Viewers Easily Spot AI-Generated Social Video Scripts

The distinct stylistic quirks of large language models are making AI-generated video scripts highly recognizable to audiences on platforms like TikTok and YouTube. According to a critique highlighted by technologist Simon Willison, viewers are increasingly able to identify when a creator has outsourced their writing to an AI assistant, pointing to specific structural patterns and an overall lack of authentic human perspective.

The critique, originally voiced by online commentator @therealcornpop, identifies several dead giveaways common in AI-generated scripts. These "AI-isms" include repetitive rhetorical structures such as "it's not X, it's Y," an overreliance on the "rule of three" for list-making, and a staccato writing style that uses excessive punctuation to mimic intellectual depth. Beyond these mechanical patterns, the primary flaw identified is a total absence of a unique personal voice or strong opinions on the subject matter.

For AI practitioners and developers, this growing audience fatigue highlights a critical challenge in natural language generation. Standard reinforcement learning from human feedback often pushes models toward a polite, neutral, and formulaic tone, which directly contributes to the recognizable patterns that viewers now reject. To overcome this, developers may need to focus on training models that can emulate more diverse, opinionated, and less structured human speech patterns.

Ultimately, the critique serves as a warning for creators relying on automation for audience engagement. Without what the post describes as a "definitive sort of spear of your voice," automated scripts fail to connect, proving that raw efficiency cannot replace genuine human perspective in digital media.