YouMind
Sign in

ChatGPT Writing Tutorial: Eliminate AI Flavor by Chatting, Not Prompting

@bbkirstry
SIMPLIFIED CHINESENov 19, 2026
103K
837
115
75
1.4K

TL;DR

This guide details a ChatGPT workflow using conversational context, user profiles, and iterative editing to eliminate AI flavor and maintain an authentic human voice.

Friends often complain to me: "AI-written content looks terrible!" and "Writing on my own is too hard!"

They ask, "Can you help me edit my article?"

I started wondering, why do we still struggle so much with removing the 'AI flavor' when AI capabilities are already this strong?

There are endless writing prompts, skills, agents, and workflows. Everyone tries their best to make AI output sound human.

The result? You generate one draft, revise it five times, and it still feels robotic.

But I find writing articles with ChatGPT very smooth. I rarely go through that process.

To be honest, it only takes me half an hour to write an article.

And I don't even have a fixed writing prompt!

Often, I just think of something interesting, open ChatGPT, and start chatting.

As we chat, I might end up with an article I'm quite satisfied with.

So today, I want to share an AI writing method that might differ from many tutorials you've seen before.

You don't need to install writing skills or study complex prompts.

You just need to open ChatGPT's Chat mode and start talking.

Note, I mean really chatting!!!

Not typing a command like "Write me an article" into the chat box.

In my view, these two things are worlds apart.

This post will be long, but I'll try to break down my workflow for using AI to assist in writing clearly.

From establishing context and enabling memory, to chatting, organizing viewpoints, drafting, and editing, I'll include actionable steps ✍️

If you're struggling with AI writing, self-media creation, or those endless "remove AI flavor" tutorials, please read carefully and save this for later!

小晚不在 - inline image

01 | We Overestimate the Barrier to Internet Writing but Underestimate How Hard It Is for AI to Understand a Person

I've always felt people have contradictory understandings of AI writing.

On one hand, we overcomplicate internet writing.

True literary writing is an art form. Rhythm, narrative techniques, imagery, subtle rhetoric, and appropriate white space all require long-term accumulation and practice.

I deeply appreciate such writing and believe words deserve to be taken seriously.

But most internet content creation first solves the problem of expression.

You attended an event, met interesting people, and want to share your experience; you accomplished something and want to tell others about it; you gained new insights on a topic and want to discuss them.

An excellent self-media creator isn't necessarily an excellent literary writer.

You might not be good at organizing text, but you have rich experiences, sharp observations, and your own judgments.

Yet many people get nervous just thinking about "writing an article." How to start? Should there be a hook? How many paragraphs? How to create emotion? Should there be a sublimation at the end?

Before they even start expressing themselves, they're intimidated by writing techniques.

On the other hand, we oversimplify AI writing.

It seems like as long as you give AI a topic and a prompt, it can produce an article that truly belongs to you.

But generating 2,000 words is completely different from understanding what a person truly wants to express.

Why do you care about this? Which details moved you? What do you hope others see in it?

Sometimes, even you haven't figured it out yet.

How could AI accurately understand all this with just one instruction?

So I increasingly feel that we overestimate the skills needed to write an internet article, but underestimate the understanding required to write an article that is truly yours.

小晚不在 - inline image

02 | Why Does Asking AI to "Write an Article" Immediately Create a Robotic Feel?

I've noticed a subtle phenomenon.

If you directly tell ChatGPT:

"Write a 1,500-word article about personal growth in the AI era. Keep the language natural and engaging, suitable for posting on X."

You might quickly get a structurally complete, fluent article.

But it will likely contain familiar robotic expressions:

  • "In this rapidly changing era..."
  • "We need to re-examine..."
  • "What truly matters is not... but rather..."

Even though you asked for naturalness, it starts giving a formal speech.

I later realized this might be related to the task we assign to AI!

When you ask it to "write an article," it tends to generate a work that fits common article formats. Introduction, transition, viewpoint, conclusion—everything is neatly arranged. But it doesn't know why you want to talk about this, nor which details are particularly important to you.

As a result, specific observations are easily replaced by standardized expressions.

What if we try a different approach?

Open ChatGPT and say:

"I recently discovered something interesting, but I don't know how to share it with others. Let's chat about it."

Then talk about what you experienced, why you found it fun, and what parts you haven't figured out yet.

It responds, you continue speaking.

It misunderstands, you correct it.

It suggests an angle you hadn't considered, and you discuss further.

Throughout the process, you both focus on that specific thing.

No one is rushing to create golden sentences or sublimate the theme.

You are simply trying to: explain a thing clearly.

I think the beauty of Chat mode is that it allows you to start communicating even when you don't yet know exactly what you want to say.

Of course, chatting can also produce formulaic language, and direct writing requests can yield good articles.

The key is whether AI has understood you through sufficient context.

My experience is that starting with a chat often preserves those unformed ideas and naturally expressive thoughts.

Many methods to "remove AI flavor" actually just replace a few words or adjust sentence structures after the article is written.

But if the author's true intent wasn't clear from the start, no amount of polishing can make the article develop its own unique flavor.

So-called "AI flavor" is sometimes a language issue, and sometimes an understanding issue.

小晚不在 - inline image

03 | Let ChatGPT Know You: Building Long-Term Collaborative Context

Okay! Enough theory, let's get practical.

Many friends wonder why ChatGPT seems to understand me so well during our chats.

Sometimes I just say, "Guess what I'm thinking," and it analyzes things I haven't even said aloud.

I think part of it comes from the model's inherent understanding ability, and part comes from the context we've accumulated over time.

It knows my work history, professional skills, interests, and expression preferences. It has seen many of my articles.

More importantly, we have had countless corrections.

When it misunderstands, I explain why it's wrong;

When it writes too verbosely, I ask it to cut;

Sometimes I even rewrite a few sentences myself to help it re-understand my meaning.

So the context worth accumulating includes not just what you said, but why you like or dislike certain things, and how you make judgments.

How can ordinary people start from scratch?

① First, Give Chat Some Background About Yourself

I suggest opening ChatGPT, starting a new normal chat, and telling it four things:

  • Identity & Experience: Who you are, what you do, and your background.
  • Interests & Strengths: What you follow, what you're good at, and what you're researching lately.
  • Expression Preferences: How you like to speak, words you hate, and articles you admire.
  • Creation Goals: Which platforms you want to share on and who you want to connect with.

For example, if you're a product manager learning AI, you can say:

"I've been a product manager for three years. Recently, I've been researching AI and doing some small projects myself. I want to start sharing my learning process, project experiences, and product thoughts on X. I have many ideas when discussing products with friends, but I'm not good at writing long articles. I hope you can gradually understand my experiences, thinking style, and language habits, and accompany me in discussing and organizing content."

Just like getting to know a new friend, let it get to know you as you interact.

② Show It Articles You've Written Before

I highly recommend this!!!

Find two articles you were satisfied with, and one you weren't (preferably written by yourself, not AI).

Then tell Chat:

"Here are some articles I wrote before. Please analyze what issues I focus on, how I organize my thoughts, and what characteristics my language has. Explain with specific sentences; don't just say I'm sincere or have a personal style."

After it analyzes, you can continue correcting. For example:

"Actually, what I like isn't the exclamation marks, but the narrative rhythm within them."

"Although this piece has a complete structure, I really disliked it because it explained a simple concept in an overly complex way."

I believe there are three important layers in expression: Perspective, Logic, and Language.

  • Perspective: What you care about and what you observe.
  • Logic: Why you think this way and your reasoning process.
  • Language: How you usually speak and express things.

Getting AI to understand these three things is more important than simply mimicking your catchphrases.

③ Enable Memory to Make Future Collaboration Easier

If you plan to use ChatGPT long-term, go to Settings > Personalization and check/enable Memory.

It helps subsequent conversations retain some long-term information and preferences related to you.

But note, Memory does not equal saving every chat completely, nor does it guarantee remembering all details.

Keep your own copies of important articles, creative materials, and project backgrounds, and provide them when needed.

You can also periodically ask Chat to re-summarize its understanding of you to see if anything is outdated or inaccurate.

After all, you are constantly growing, and past preferences may not fully support your current expression.

💡 Practical Step for This Chapter:

Today, open a new Chat, introduce your basic background, and provide two satisfying articles and one unsatisfying one.

Then ask:

"What kind of communicator do you think I am? What are the characteristics of my focus points, thinking style, and language habits? Please judge based on specific content, and tell me directly if there are areas you don't understand."

Correct any inaccuracies immediately and chat until you establish your own stylistic logic and judgment criteria.

小晚不在 - inline image

04 | Start Really Chatting: Discover Worthwhile Topics from Genuine Feelings

Next is my favorite part.

Please forget about "writing an article" for now.

A few days ago, I tried teaching my method to a friend named Xiaokai.

At first, I asked him: "What do you want to express?"

Later, I realized this question is actually hard to answer.

Because if a person can already accurately tell me what they want to express, the rest of the process would likely be much simpler.

So I changed the question to:

"What did you feel? What was fun? What kind of event was it? And what did you think was important?"

See, isn't that easier to answer?

For example, if you recently attended an event and met interesting people; or encountered a strange problem at work and suddenly gained new insight.

You can just open Chat and say:

"I recently experienced something quite interesting and want to chat about it. I haven't fully figured it out yet, so listen to me first, and let's see together what's worth sharing."

Then just chat normally.

Typing or voice input works fine.

You can even copy chat logs, meeting transcripts, or scattered notes you want to discuss to help supplement the context.

If you still don't know what you want to say after chatting, ask:

  • "What do you think is interesting about this situation?"
  • "After listening, what are you most curious about?"
  • "What do you think I cared about most in what I just said?"

Then carefully look at its response.

If it hits the mark, keep talking; if it misunderstands, refute it directly.

For example:

  • "It's not that profound, I just thought it was funny."
  • "That judgment is too absolute, I didn't mean to express that."
  • "You missed an important detail, let me add it."

Or:

"Your second point was really interesting, let's continue discussing that!"

This is exactly how I use ChatGPT!!!

And surprisingly, sometimes when it misunderstands, my urgency to correct it actually helps me realize what I truly want to say.

A few days ago, I originally just wanted to discuss tools like GrokBot, allowing solo companies to use multiple bots collaboratively.

But Chat asked me: Do you care more about increased execution capability, or about humans being able to step back from execution to focus on judgment and creation?

I answered, both, but the latter more so.

By this point, our discussion extended from AI tools to how humans should reorganize their work.

You see, that complete viewpoint initially emerged slowly during the discussion process.

Don't treat AI's understanding of you as the standard answer. Treat its responses as material for continued thinking.

Also, you don't need every chat to produce an article.

Some chats might just help you clarify one thing, and I think that's great too!

💡 Practical Step for This Chapter:

  • Find a small recent event that triggered a reaction in you, and chat about it for ten minutes.
  • Don't write an article yet; see if you can articulate more details and gain new understanding.
  • Finally, ask it to summarize what you cared about most, and personally correct at least one inaccuracy.
小晚不在 - inline image

05 | From Chat to Article: Define Readers, Explicit Purpose, and Implicit Purpose

Okay, suppose you've chatted happily and finally feel:

"Yes!!! This is exactly what I want to talk about!"

What's next?

I suggest considering a very important internet "communication" question:

Who is this article specifically for? What do you hope it accomplishes?

Especially for friends building a personal IP long-term, I strongly recommend distinguishing between the explicit and implicit purposes of the article.

① Explicit Purpose: What Value Do You Provide to Readers?

Take this AI writing tutorial we are currently writing as an example.

My target readers are those who have started using AI, want to share content online, but still find writing difficult.

So the explicit purpose is to help everyone correct cognitive misconceptions and learn to use ChatGPT to complete an article that is truly theirs, starting from chatting.

Therefore, I must provide specific operational steps, cases, and judgment criteria.

If I spend half the time praising how powerful AI is, but readers still don't know how to operate, then this tutorial fails in my eyes.

② Implicit Purpose: How Do You Want Readers to Perceive You?

I also hope that through this article, people gradually understand how I think.

For example, I love researching new things, enjoy free discussions with AI, and care deeply about human feelings, aesthetics, and creativity.

**Personally, I hope you see not just "Xiaowan knows how to use ChatGPT to write articles."

But also understand my observations and judgments regarding AI, content creation, and the relationship with humans, and truly derive inspiring information and strength from it.**

The explicit purpose determines why readers are willing to watch; the implicit purpose determines how they remember you after reading.

Of course, this isn't about forcing self-promotion in every article.

If you're a designer, let your work and aesthetic sense introduce you;

If you're a product manager, prove your abilities through project experience and product judgment.

What you've done, what you care about, and why you judge things this way inherently allow readers to gradually recognize what kind of person you are.

③ Hand These Goals to Chat and Discuss Structure

You can directly say:

"I want to turn what we just discussed into an article. The target readers are... Their main problem is... The explicit purpose is to help them understand... and learn... The implicit purpose is to help them gradually recognize my thinking and capabilities in... aspects. Based on the viewpoints and real cases we discussed, please help me organize the article structure first. Clarify what each section needs to cover and which details to keep. Do not write the body text yet."

If you don't know your implicit purpose yet, you can ask Chat to propose a few possible directions for you to judge.

But note, do not fabricate experiences, exaggerate abilities, or force sublimation just to build a personal IP.

💡 Practical Step for This Chapter:

Answer these three questions:

  • Who is this article for?
  • What will readers gain after reading?
  • What impression do you hope readers form of you through this article?

Then ask Chat for a structural proposal and check if each part serves a practical purpose.

小晚不在 - inline image

06 | Let Chat Organize the Article: Determine Structure Before Writing Body Text

When I write articles, I like to discuss the overall logic with Chat first.

Because a single chat might produce many interesting viewpoints, but that doesn't mean every viewpoint needs to be included in the same article.

Some things are worth emphasizing, some content is repetitive, and some information is known to you but not to the reader—all need reorganization.

So I ask:

"If we were to share what we just discussed with someone who doesn't know me at all, how do you think it should be arranged?"

Let it organize chapters, key points, cases, and order first.

Then judge.

  • "These two sections are actually saying the same thing, merge them."
  • "Too much explanation here; readers should have understood it by now."
  • "This example is important, don't delete it."
  • "The title of this section should directly tell readers what they will learn."

Once the structure is basically satisfactory, then ask it to draft the initial version.

Most of the time, I say this:

"Based on what we just discussed, please explain the whole story and my viewpoints clearly, as if sharing with a friend. Try to express it from my perspective, logic, and language. Retain real experiences, specific details, and emotions. Do not add judgments I haven't expressed, and do not fabricate facts just to make the article seem complete."

If writing a tutorial, you can add:

"Assume the reader has zero experience. Help me check which operational steps I described too vaguely, where others might not know what to do."

This point is especially important.

Things you are already very proficient in might be exactly the parts ordinary people don't know how to handle.

So you need to break down those habitual experiences into actions others can understand and execute.

Of course, don't treat the structure as something absolutely unchangeable.

During the writing process, you might discover new viewpoints or even change your original mind.

Then continue chatting and adjusting.

Writing the article itself is part of continuing to think.

💡 Practical Step for This Chapter:

  • First, let Chat output the structure, then check the content value, cases, and expression order of each chapter.
  • Confirm satisfaction before generating the body text.
  • If unsatisfied, solve structural problems first; don't rush to repeatedly rewrite the entire article.
小晚不在 - inline image

07 | Editing: Learn to Judge Specific Issues, Don't Just "Remove AI Flavor"

I think this is a step worth practicing seriously.

Many people feel AI-written articles are unnatural, so they repeatedly request:

"Remove the AI flavor."

"Make it sound more human."

"Make the language more vivid."

But where exactly is it unnatural?

Sometimes AI amplifies your emotions, sometimes it deletes real details, and sometimes it summarizes a conclusion you don't agree with.

So, I suggest checking from three places.

① Check Viewpoints: Are They Truly What You Wanted to Say?

Did the article misunderstand your meaning?

Did it turn a personal experience into a universal law?

Did it add viewpoints you never mentioned?

If the content direction is wrong, beautiful language is useless.

② Check Structure: Can Readers Follow Along Easily?

Is there missing background information?

Is there repetitive explanation?

Does a section drag on without providing new information?

Don't repeat things that are already clear.

③ Check Language: Would You Actually Say This Sentence Yourself?

I care about this point especially.

Would I really say this to a friend in daily life?

If not, I change it.

A few days ago, I helped my friend Xiaokai edit an article about the Guancha Conference.

He attended a three-day event, made many friends, and experienced various products, so he originally had rich material.

But some AI-polished parts covered his originally vibrant expressions.

So I first discussed with ChatGPT which content was worth keeping and where it felt unnatural, then adjusted the originally scattered nine sections into five.

There was a section Chat thought could be kept, but I felt the emotion was off, so I wrote my own version:

"In this home, warmth flows between us.

You give me feedback, I improve the product for you, back and forth, we care for each other like this."

What I like is this specific, mutual caring relationship between people.

Moreover, Xiaokai originally had many cute expressions.

For example, seeing the product red/black list, he wrote:

"I died laughing, Claude was almost drowned in red crosses!!!"

When he didn't get a group photo, he wrote:

"Ahhh, what a pity no group photo!"

Do these phrases have deep writing techniques?

Not really.

But you can intuitively feel this person's emotions and feelings.

I think Xiaokai actually possesses excellent expressive ability.

I just need to discuss with GPT to reorganize the valuable parts.

So if you feel AI wrote something incorrectly, tell it directly what's wrong, or even write a sample sentence yourself.

"This sentence isn't what I meant."

"I wasn't that moved at the time."

"This paragraph is too wordy."

"The first two paragraphs are good, keep them all; only modify the third."

Don't casually ask AI to rewrite the whole thing.

The more specific the feedback, the easier it is for Chat to understand your meaning.

💡 Practical Step for This Chapter:

  • Find an AI article you are dissatisfied with.
  • Point out one viewpoint issue, one structural issue, and one language issue separately.
  • Ask Chat to modify only the corresponding content, then read through to confirm.

Remember: Every revision should solve a specific problem, not just find a nicer-sounding phrase.

小晚不在 - inline image

08 | Long-Term Collaboration: Let Chat Understand You Better, and Let Yourself Express Better

The methods above should be enough to help you complete your first article.

But if you want to create content long-term, I think you need to cultivate a habit:

After every creation, look back at what you actually learned.

For example, if during this edit you realized you particularly like retaining authentic chat tones,

next time, continue telling Chat this requirement.

Or if you notice you frequently dislike lengthy explanations or low-information transitions, ask it to help organize these with specific examples.

Reader questions, post-publication feedback, and new difficulties encountered in the next creation can all be discussed with Chat during the process.

For example:

  • "In the articles we've written recently, which expressions do I particularly like?"
  • "What issues have I repeatedly corrected? Do these issues share a common cause?"
  • "What do you think I'm currently best at sharing? Where do I still need practice?"

When I previously wrote a tutorial on starting a Xiaohongshu account, I emphasized "learning while doing."

Start publishing first, then gradually build content judgment through actual feedback.

AI writing is the same.

You don't need to master all techniques from the start, but you should become clearer about your own thoughts through repeated expression and revision.

One more thing I want to remind everyone:

Don't fix yourself into a permanent persona just so AI remembers your style.

Today you might like lively, humorous expressions; tomorrow you might want to write a quiet, elegant article seriously. You change, your aesthetics grow, and your focus shifts.

AI can help you understand yourself, but don't let it decide who you should be.

💡 Practical Step for This Chapter:

  • After finishing an article, give Chat both the draft and the final version.
  • Ask it to summarize the three most important changes in this revision, what should be retained in the future, and which errors are worth avoiding.
  • Organize truly long-term effective preferences into your own creative reference materials.
小晚不在 - inline image

09 | AI Makes Expression Easier, But Don't Forget Life ❤️

Last night, I talked with friends in a group about anxiety caused by AI.

Someone said they haven't written anything without relying on AI for a long time.

Someone said they find it hard to concentrate on reading physical books lately.

Another friend said that although AI seems capable of everything, many important details still need manual handling once you enter actual business scenarios.

These words reminded me of myself in October.

During that period, I often felt my brain jumping around the internet. There were too many new tools, features, projects, and ideas every day.

But when I walked home after work, chatted with friends again, and felt the real life around me, I realized the distance between online information and reality.

Later, I started adjusting my pace.

Sleep more, read more, chat with friends more, watch movies, listen to music.

I gradually realized that some things really cannot be rushed.

For example, if you want to write a video script with genuine emotion, many details require stopping to recall slowly. You need to re-feel those moments that truly moved you.

These things won't be better just because AI can generate more content per second.

So I especially want to tell anxious friends:

You don't need to chase new technologies every day to prove you aren't falling behind.

AI can help you accomplish many things, but life still requires you to experience it personally.

And I increasingly feel we need to re-understand: How should a person's ability be measured?

Previously, a person might have many interesting ideas but couldn't bring them to fruition due to poor writing, design, or technical skills.

Now, AI opens new possibilities.

You can use it to learn unfamiliar knowledge, try things you couldn't do before, and organize those unclear ideas through dialogue.

This is why I love chatting with ChatGPT so much.

Sometimes we chat out an article.

Sometimes we chat out a project.

And sometimes, it just helps me understand a problem I couldn't figure out before.

Many times our conversations don't produce new works or articles, but being able to research fresh things and discover interesting connections anytime makes me happy.

Perhaps what is most worth releasing in the AI era is not productivity, but those ideas that were never expressed because of lack of proficiency, professionalism, or confidence.

So I hope we don't forget what we originally wanted to do just because AI can do everything.

Go read, meet friends, live well.

Allow some things to happen slowly, allow yourself to not know the answers temporarily.

Some things worth writing down might be hidden in that afternoon when you put down the computer and walk outside.

When you truly have something to say, come back and open ChatGPT.

I believe you can explore happily with it and create some wonderful content 🍻

Above are all the tips I use to quickly organize fragmented ideas into articles using ChatGPT's Chat mode. If you have questions, welcome to leave comments!

I really hope more people can experience this joy!!

Many people have things worth expressing, they just haven't found a sufficiently relaxed way to write them out.

Hope this article helps you finish the last mile on your writing journey 🫶

———

This is @XiaowanBuZai 🙋

Practitioner of content marketing in the AI era, loves researching fun content, fun humans, and the new possibilities brought by AI.

My thinking assistant Skill "Cyber Xiaowan":

https://github.com/cyberxiaowan/cyber-xiaowan

My long-image layout Skill:

https://github.com/cyberxiaowan/xiaowan-wechat-layout-skill

This article comes from my actual content creation experience, as well as recent discussions with ChatGPT, friends, and community members. Hope it helps everyone find a more relaxed and suitable way to express themselves! If you need to join my paid Q&A group, feel free to DM me~ 💌

Remix in YouMind

Turn one viral article into a full content workflow

Collect the source, decode the pattern, create assets, draft the story, and distribute from one AI workspace.

Explore YouMind
For creators

Turn your Markdown into a clean 𝕏 article

When you publish your own long-form writing, images, tables, and code blocks make 𝕏 formatting painful. YouMind turns a full Markdown draft into a clean, ready-to-post 𝕏 article.

Try Markdown to 𝕏

More patterns to decode

Recent viral articles

Explore more viral articles