WeChat Article Prompt Builder

Structure is what turns an AI draft into a publishable WeChat article

When you ask a large language model — DeepSeek, Qwen (通义千问), Doubao, Kimi, ERNIE or Zhipu — to write a WeChat (公众号) article, the quality of the draft depends far more on how you brief it than on which model you pick. A vague "write me a 公众号 article about productivity" yields generic filler; a structured brief that names the topic and headline angle, the one or two core points, the exact target reader, the article structure, the voice, the length and the keywords yields a draft you can actually publish. This builder keeps that structure for you: fill the fields, and it joins them into a clean writing prompt with a leading instruction line followed by clearly headed sections — topic, core argument, reader, structure, voice, length, keywords — each prefixed with a Markdown-style heading the model can read at a glance. The result is the kind of brief a seasoned 公众号 editor would hand a writer, only assembled in seconds.

The two highest-leverage fields are the topic / headline angle and the article structure. A specific angle — "for first-time managers, why one-on-ones fail and how to fix them" — steers the whole piece, and asking for several candidate headlines up front saves you a separate round of prompting. The structure field is what gives a WeChat article its rhythm: spelling out a hook to open, a few sub-points each backed by a concrete example, a memorable punch line, and a clean close stops the model from returning one undifferentiated block of text. Add a core argument so the article has a backbone, and the draft stays on track instead of wandering across half a dozen loosely related ideas.

Reader, voice and length are what make it sound like you

The fields people skip and regret are the target reader, the voice and the length. Naming the reader — a career beginner, a busy parent, a fellow engineer — keeps the depth, the examples and the wording pitched correctly, because an article written for everyone is read by no one. The voice ("warm, professional, a little internet-savvy") is what keeps a series sounding like the same human author rather than a different model each week. And a target length — say 1,500 to 2,000 字 — matches how people actually read on WeChat: long enough to be worth opening, short enough to finish on a phone over coffee. None of these constraints limit the model; they focus it on the article you actually want.

Because the output is a structured plain-text brief, the same prompt is portable across every major Chinese model and works just as well on ChatGPT, Claude or Gemini. Treat the first draft as exactly that — a draft. Run the prompt, read the result, and notice where the article drifts: too generic means tighten the core argument; wrong tone means sharpen the voice; thin examples means ask explicitly for a concrete case per sub-point. Two or three rounds of that usually turn a mediocre first pass into a piece you are happy to publish. And because the whole tool runs locally in your browser, you can iterate freely — tweak one field, copy again, and test — without anything you type ever leaving your device, being sent to a model, or being stored. Always read and fact-check the draft before publishing, and fold in your own real examples so the final article keeps a genuinely human voice.