📄 中文摘要
随着公司在文档平台中集成 AI 助手，测试团队的成员发现 AI 在回答用户问题时的表现超出了预期。这一过程不仅关注 AI 的准确性，更涉及文档在大型语言模型（LLMs）中的表现。测试中发现，LLMs 的非确定性特征并不是缺陷，AI 有时会对同一问题给出不同的答案，这引发了对技术写作方式的重新思考。文档的结构和内容需要适应 AI 的处理方式，以提高其在 AI 环境中的有效性。
为人类写作已不再足够，面向 AI 的写作成为工作的一部分
📄 English Summary
Writing for humans is no longer enough. Writing for AI is now part of the job
The integration of an AI assistant into a documentation platform led to unexpected insights regarding technical writing. The testing team aimed to evaluate how well the AI could respond to user inquiries using existing documentation as a reference. The exercise highlighted that the accuracy of AI is not the only concern; it also revealed how documentation interacts with large language models (LLMs). One notable observation was the non-deterministic nature of LLMs, which sometimes provided different answers to the same question. This variability prompted a reevaluation of technical writing practices, emphasizing the need for documentation to be structured in a way that enhances its effectiveness in AI contexts.