Study finds 40% of US desk workers receive AI-generated slop
A joint study by Stanford and BetterUp found that 40% of U.S. desk workers received low-quality AI-generated work in the past month, costing recipients nearly two hours. Experts propose frameworks to distinguish high-quality AI assistance from generic output, emphasizing human control over narrative architecture.
A joint study by Stanford and BetterUp found that approximately 40% of U.S. desk workers received work products considered 'sloppy' in the past month, costing recipients close to two hours of time. A separate Glean survey of 6,000 workers indicated that nearly 70% shipped work without reading it before sending it to others. Deezer reported that more than half of the music uploaded to its platform daily is fully AI-generated, yet AI-generated tracks account for only 1% to 3% of total downloads. The term 'slop' was named Merriam-Webster's word of the year for 2025, defined as low-quality digital content produced in quantity by AI. Analysts argue that 2026 marks a shift where agentic tools produce documents and code in volume, moving the cost of processing generic output from the creator to the receiver. To address this, a strategic framework has been proposed to distinguish high-quality AI-assisted work from 'slop.' Key techniques include thinking before opening the tool, explicitly stating confidence levels to force model pushback, and defining the audience and desired outcome upfront. During execution, users should request multiple distinct directions rather than accepting the first average answer and pre-define strategic interception points for review. A 'board' of AI personas, including a skeptic, expert, and devil's advocate, can review drafts. Figma has formalized rules for AI writing, including labeling AI-generated content. A NIPS paper noted that different models converge on similar answers, meaning switching vendors does not ensure originality. The framework emphasizes that human point of view and narrative architecture must remain with the human, while AI handles research and drafting. Hosts argue that originality is lost when users outsource thinking to AI, advocating for a workflow where humans retain control over perspective and structure.