AI can accelerate writing, but the first draft is not the right unit of efficiency for evidence-heavy content; defensible finished content is. In Noy and Zhang’s Science experiment with 453 professionals, ChatGPT reduced average completion time for midlevel writing tasks by 40% and increased evaluator-rated quality by 18%. However, if an editor must investigate statistics, citations, product claims, and conclusions, part of that gain disappears. The article proposes a method: Evidence → Claims → Reasoning → Prose → QA, avoiding asking one prompt to research, infer, judge, and write simultaneously. This proprietary approach, designed by Egor Kaleynik, is not presented as a scientifically validated universal workflow. AI slop comprises two problems: epistemic slop (fabricated facts, nonexistent citations, outdated information, claims beyond sources) and editorial slop (generic, bloated, repetitive content). These require different fixes. The 2026 Nature paper “Synthesizing scientific literature with retrieval-augmented language models” reports that GPT-4o fabricated citations in 78–90% of cases in their evaluated citation task, while the retrieval-based OpenScholar system improved correctness and citation performance. A common AI-writing process—topic, outline, draft, fact-check, rewrite, publish—feels efficient but creates verification debt. The article was produced using the same workflow described.