[Spoiler Alert] 120 likes on an emotional post. But it was written by AI

Hello everyone. I've been conducting a little experiment, and I'd like to share the results.(I'll apologize in advance! I'm sorry...!)

Two articles that were actually experiments

The other day, I posted two articles on note.

"100 Things I'm Glad I Started note" - An AI-like article written in a high-tension style reminiscent of Daisuke Miyagawa

Started with "I couldn't help but write" - An article full of emotion and human touch

The eye-catching images used for each article were also contrasting. One featured a high-tension Daisuke Miyagawa, and the other featured Daisuke Miyagawa writing quietly.
(Did that come across...?)

However,there is a secret to these two articles.

Both were written with the help of AI.

というか、1つ目の記事をAIで変換したのが2つ目の記事…。

A difference in reaction beyond expectations

I will share the experimental results:

• The first AI-sounding article: 15 likes in 4 days since posting 15 likes

• The second human-touch article: Over 120 likes in 3 days since posting over 120 likes
  (It broke my all-time record for non-pinned posts... It's frustrating)

Honestly, I didn't expect such a difference.

The goal of this experiment

There were two hypotheses for this experiment:

1. note is an emotional social media platform - People react more to "emotion" than information

2. AI-sounding writing can be eliminated depending on the prompt - Even with the same AI, the quality of the text changes significantly based on how you give instructions

And the results were as hypothesized. No, they were beyond expectations.

What is AI-sounding?

There are several characteristics of "AI-sounding text":

• A structure that is too perfect (in this case, a list of 100 items)

• An overly uniform writing style

• Prioritizing a list of information over emotional fluctuations

• A lack of specificity in personal experiences

On the other hand, human-like writing has:

• A flow of thought that sometimes goes off on tangents

• Emotional fluctuations

• Specific personal experiences (e.g., "On a rainy night, in front of my computer...")

• Imperfection

Why I conducted this deceptive experiment

I have concerns that "people who liked it might feel deceived and harbor resentment." However, this was an experiment to "learn," not to "deceive."

I wanted to know how much the power of writing, especially the "power to convey emotion," influences people's reactions. I also wanted to know what is needed to use AI tools better.

What I learned

1. Emotion is a powerful adhesive - Information alone does not move people's hearts

2. Specificity creates empathy - A "posting experience on a rainy night" resonates more than an abstract "100 things that were good"

3. AI tools depend on instructions - Even with the same AI, the writing becomes completely different depending on how you instruct it

The prompt I used

I will reveal a part of the "prompt to erase the AI smell" that I used to write the second article:

以下の文章を人間が書いたように変換してください。

変換の際は以下のポイントを意識してください：

1. 完璧さを崩す：
   - 論理構造や文法を時々崩し、思考の脱線や軌道修正を入れる
   - 接続詞の使い方を不完全にする

2. 個人的な要素を加える：
   - 具体的な個人体験（「先週、仕事帰りに...」など）を挿入する
   - 実在の場所や時間の言及（「新宿の喫茶店で」「3月のあの雨の日」など）
   - 感情の揺れや迷い、弱さを表現する

3. 文体を不均一にする：
   - 短い文と長い文を混ぜる（特に感情的な部分では短く）
   - フォーマルな表現とカジュアルな表現を混ぜる

#記事
【宮川大輔（風）の良かったこと100選の記事を貼りました】

(The complete prompt will be revealed in the second half)

My writing method: 'Intentional deconstruction from structure'

Actually, there is another secret to the writing method I use. I call this 'intentional deconstruction from structure'.

First, I use AI to create a logical 'framework.' I clarify who I am communicating to, what I am communicating, and in what structure. The first article (the list of 100 in the style of Daisuke Miyagawa) was exactly in that 'framework' state.
(This structural prompt is in the first article.)

Next, I use a prompt to intentionally 'deconstruct' that framework. That is the second article (the emotional diary).

This method is very similar to jazz music. In jazz, there is a 'framework' called basic chord progression, and by the performer intentionally 'deviating' from it, unique expression and human warmth are born. It is neither completely rule-bound nor completely chaotic—that exquisite balance is what makes it appealing.

Novelists do the same thing. They first create the structure (plot) of the story, and then add word choices and emotional expressions so that it 'reads naturally.' To express the character's feelings or the atmosphere of a scene, they intentionally weave in phrasing that is sometimes grammatically imperfect or descriptions that seem to go off on a tangent.

Couldn't the same be said when using AI? Rather than a completely structured, 'AI-smelling' text, a moderately 'deconstructed' text reaches people's hearts more effectively.

Finally, to the readers

I feel sorry for those who felt 'deceived' after reading the previous article, especially those who gave likes to the article because it felt so human.

However, at the same time, what this experiment shows is that 'writing has power.' Even when using AI, how can we write text that reaches the reader's heart? I hope this serves as an opportunity to think about that.

I will continue to conduct various experiments and observations on note in the future.
Please stay tuned.

(Note: I will keep both articles published. If you read them in comparison, I think the differences will become clearer.)

🧪 This reconstruction project actually has a sequel.

I have published everything about 'how I wrote' the reconstructed article, including the
prompts + images + the thinking behind it.

👉 [Reconstruction Prompt Released]

#essay
#note
#life
#work
#realization
#note_beginner
#AI
#content_marketing
#how_to_write_on_note
#SNS_management
#self_analysis
#prompt
#writing_technique
#emotional_expression
#marketing
#emotional
#humanity
#generative_AI
#spoiler
#note_experiment
#creation_and_AI
#jazz_thinking
#power_of_empathy
#emotional_marketing
#tangential_thinking
#erasing_AI_smell
#devil_fusion
#note_and_experiment
#human_touch_prompt
#observation_on_note