It is very necessary to establish a strong disabled vocabulary in order to completely break away from the machine flavor of AI output. I have helped organize a more comprehensive blacklist, which can be directly added to your prompt template. Feel free to bookmark it 1. Directly block these nonsense word libraries Forcefully add this sentence at the end of your prompt: It is strictly prohibited to use "in short, not only that, the key is that, in-depth exploration and opening up new chapters, not only that, but more importantly, it is worth noting that it is a powerful tool. In depth exploration, unlocking potential, highly attractive, amazing, comprehensive, multi-dimensional, the key is". These words are the fingerprints of AI. Delete them and strictly prohibit the use of double quotes and dashes. 2. Accurately identify the audience's specific concerns and construct a precise audience portrait Don't say it's for newcomers in the workplace, you should say it to the operations guy who hasn't finished his weekly report by 3 pm every day and wants to leave work early to go to the gym. When your audience portrait has specific pain points, the cases selected by AI will no longer be empty truths, but life-saving practical goods. If you only ask AI to write an article about financial management, it will only give you a bunch of big truths. Try to concretize your audience: Your reader is a senior product manager who has worked in Shanghai for five years, earns 20000 yuan per month but still feels anxious. He dislikes listening to big truths and only cares about how to achieve a stable annual return of 10% through quantitative trading. When AI has a specific communication partner, its selected cases and tone will naturally become down-to-earth, and the warmth of a real person will immediately come out. 3. Mandatory requirement for non-linear expression AI prefers to use first, second, and last by default, and this layout is too much like an instruction manual. Try to tell it: prohibit the use of ordered lists, use more short sentences, and add a bit of colloquial inflection, such as saying it's plain or not that mysterious. 4. Suggested operating instructions You can add this paragraph directly at the beginning of the prompt: Please strictly follow the following rules when writing: It is strictly prohibited to use any grand opening remarks and directly enter the theme. Delete all concluding statements. Replace all adjectives with specific verbs or facts. Prohibit the use of parallelism and overly structured confrontations. 5. Final adjustment: Don't expect AI to produce a perfect score on the first attempt. After it generates the initial draft, you can tell it: the current content looks too much like what a robot wrote. Please delete all meaningless foreshadowing and summaries, and further refine the operation steps mentioned in the second paragraph by three times, adding specific code logic or failure cases. This targeted exploitation can force AI to retrieve deeper training data instead of just spinning around on the surface By following this logic, you will find that the generated text lacks the difficult to remove fatherly flavor.