ChatGPT、Google Gemini 及其他生成式人工智慧工具已向大众普及。为了获取更高品质的结果，使用者会采用提示词工程。例如，指示 ChatGPT 扮演一位好奇、爱发问的 10 岁孩童，有助于批判性地评估想法。巴莱多定律（即 80-20 法则）指出，20% 的核心资讯可以帮助使用者掌握某个主题 80% 的内容。此外，使用者还可以透过应用程式利用手机相机以相片进行提示，或上传档案。
其他提示技巧则专注于改变视角、设定限制或整理数据。使用者可以要求以 Steve Jobs 或 Winston Churchill 等名人的风格进行回复，或是要求符合特定格式的答案，例如包含 1970 年代经典歌曲的 2 小时播放清单。此外，使用者可以限制输出长度，例如要求使用 7 个字元或更少字元的单字，或者要求聊天机器人针对特定受众解释主题，这灵感来自于将概念解释为 5 个不同水平的影片系列。
进阶互动包括指示模型为 Dall-E 和 Midjourney 等其他系统生成输入、输出 ASCII 艺术或建立文字冒险游戏。使用者可以贴上外部文件以进行翻译与摘要、提供范例进行少样本学习，或者要求 ChatGPT 先提出澄清问题。最后，以表格形式要求输出可让数据轻松汇出至 Microsoft Excel 等试算表软体，而要求机器人忘记过去的互动则能确保输出不具偏差。
ChatGPT, Google Gemini, and other generative artificial intelligence tools are widely available to the public. To extract higher-quality results, users employ prompt engineering. For instance, instructing ChatGPT to act like an inquisitive 10-year-old helps evaluate ideas critically. The application of the Pareto principle, or the 80-20 rule, posits that 20 percent of the core information can help a user learn 80 percent of a given topic. Additionally, users can utilize their phone camera via the app to prompt with photos or upload files.
Other prompting techniques focus on altering perspectives, setting constraints, or organizing data. Users can request responses in the style of famous figures like Steve Jobs or Winston Churchill, or ask for answers matching specific formats like a 2-hour playlist of 1970s classics. Furthermore, users can restrict the length of outputs, such as requesting words of 7 characters or fewer, or ask the chatbot to explain subjects tailored to specific audiences, inspired by video series explaining concepts at 5 distinct levels.
Advanced interactions include instructing the model to generate inputs for other systems like Dall-E and Midjourney, outputting ASCII art, or creating text-based games. Users can paste external documents for translation and summarization, provide examples for few-shot learning, or ask ChatGPT to ask clarifying questions first. Finally, requesting outputs in tabular format allows data to be easily exported to spreadsheet programs like Microsoft Excel, while telling the bot to forget past interactions ensures unbiased outputs.