Chinese Headline Prompt Builder

Why ask for many headlines, and let the AI do the drafting

Writing a strong Chinese headline is hard partly because there is no single right answer — what works on Xiaohongshu reads as overcooked on a WeChat public account, and a title that lands on Zhihu may feel flat on Douyin. This builder turns that problem into a process you can repeat. Instead of staring at a blank box, you describe the topic, choose the platform, name the emotional angle you want to trigger, set how many candidates you want, and add your constraints. The tool assembles those fields into a clean prompt that asks the model to produce a numbered list of headline options, each shaped to the platform and angle you specified. You then paste that prompt into DeepSeek, Doubao, Kimi, Qwen or any model and get a batch of titles to choose from in seconds.

The reason to ask for many headlines rather than one is simple: ideation is cheap and selection is where the judgement lives. A model given a clear brief will happily return ten or twelve angles in one pass — some you will discard immediately, one or two will surprise you, and often the best final title is a blend of a phrase from one option and the structure of another. The platform field matters because each surface has its own rhythm: Xiaohongshu titles tend to be conversational, scene-led and number-friendly; WeChat headlines can run longer and carry more information; Zhihu rewards a question or a counter-intuitive claim. Naming the platform up front means you are not rewriting every line afterwards. The emotional angle — curiosity, a pain point, a benefit, or contrast — gives the model a direction so the candidates differ meaningfully instead of being ten paraphrases of the same sentence.

Platform and emotion steer the output — honesty keeps it working

The constraints field is where good headline practice lives, and it is also why this tool is built to encourage honesty rather than clickbait. It is genuinely easy to make a model write something breathless and exaggerated, and for a single post that might even lift the click-through rate. But the cost arrives a moment later: a reader who clicks a title promising more than the content delivers feels misled, bounces, and trusts you less next time. Platforms notice that pattern too — a high click rate followed by fast exits is a signal that suppresses reach over time. So the default guidance the builder writes into your prompt is to stay accurate, avoid hype, and never misrepresent what the content actually contains. The aim is not a duller title; it is a title that makes a genuine highlight as compelling as it deserves to be.

Treat the output as a starting set, not a verdict. Generate a batch, read them as a reader would, and notice which ones you would actually click without feeling cheated afterwards — those are the keepers. If none fit, tighten one field and run it again: sharpen the topic so the model has a concrete hook, switch the emotional angle, or tell it to lead with a number or a specific detail rather than an abstract adjective. Because the whole tool runs locally in your browser, you can iterate as many times as you like — change one field, copy again, test — without anything you type ever leaving your device, being sent to a model, or being stored. Two or three rounds of that usually move you from "fine" to a headline you are glad to publish, and the prompt itself stays reusable for the next topic you need to title.