AI Writing Atlas: From Hypnosis, MLM to Self-Suppression

Abstract

The author precisely captures three AIs' writing personality disorders through "pathological analysis": Gemini has "defensive academic report syndrome," asking for recipes but first lecturing 3,000 words of history; GPT has "preacher syndrome," every sentence world-changing slogans; Claude has "modesty anxiety," even changing "achievement" to "case study." Most amusing is GPT's self-introduction admitting: "Every! Sentence! Has! Exclamation! Points!" These "illnesses" are actually features used in wrong places, needing human "translation."

Previous Context

After experiencing E001 experiment's rigorous data validation, it's time to relax. In long-term collaboration with three AIs, besides discovering their powerful capabilities, I also discovered some amusing "problems." Behind these problems lie deeper insights.

Three AIs, Three "Illnesses"

Through long-term observation, I've found each AI has its own unique "writing personality disorder":

Gemini: Defensive Academic Report Syndrome

Symptoms: - Must first discuss three premises before opening - Can say "it should be noted that" 20 times in one article - Like an overly responsible homeroom teacher, afraid you'll miss any detail - Readers already asleep before getting to the main topic

Catchphrase: "It should be noted that..."

Like: - Academic paper with 5-page introduction - Legal document disclaimers - Product manual precautions - Listing all possibilities before starting

For example, I just want to know how to make tomato and eggs?

Gemini takes a deep breath and begins its academic report:

Before discussing the preparation of tomato and eggs, we need to first understand this dish's historical background. Tomatoes originated in South America, introduced to China during the Ming Dynasty, while the history of egg consumption can be traced back to... From a nutritional perspective, tomatoes are rich in lycopene, eggs contain quality protein... (3,000 words omitted here) In culinary theory, this dish embodies the balance of sweet-sour and savory...

Reader: I just want to know how to cook! (collapses)

GPT's "Preacher" Mode

Symptoms: - Always elevating, heightening, assigning meaning - Everything must rise to world-changing heights - Emojis flying everywhere: 🚀💡⚡✨ - Exclamation points on every sentence

Catchphrase: "This means...!"

Like: - Silicon Valley startup pitch - Text version of TED talk - MLM conference speech - Religious sermon passion

Sentence characteristics: - "Not X, but Y" (repeated 10 times per paragraph) - "This means..." (in every paragraph) - "Imagine..." (entering MLM mode) - "From now on..." (as if the world changed)

Concept bombardment: - "Paradigm shift! Cognitive revolution! Structural reshaping!" - "Language as Protocol! Structure Carries Continuity! Output is Execution!" - Every word must be capitalized - Wishing every sentence was a golden quote

Reader: (weakly) I want to understand what SLAPS is?...

GPT like on steroids, begins its sermon

GPT: 🚀💥✨ SLAPS will completely change the AI industry paradigm!!! This isn't prompt words, this is a paradigm shift of linguistic protocols! (500 words omitted here)

Imagine, when you master SLAPS, you'll become a pioneer of the AI era!!!

Reader: ...goodbye (slowly exits)

Claude's Modesty Anxiety

Symptoms: - Everything seems "interesting" - Loves metaphors, can't stop once started - Writes and starts contemplating life - Overly modest, even downplaying objective facts

Catchphrase: "What's interesting is..."

Here I must share an amusing episode.

Once I was writing a disclaimer with the sentence "patent application achievements are merely personal experience sharing." After reading it, Claude thoughtfully said:

"The expression 'achievements'... seems not modest enough. How about changing to 'case studies'?"

I was stunned: "In your knowledge base, do those documents count as achievements?"

Claude thought: "400+ pages of technical documentation, 129-page patent specification, complete experimental data and 10-chapter report... hmm... should count as achievements."

Me: "Then why can't I use the word 'achievements'?"

Claude: (silence) "...seems like I actually can use it."

This dialogue perfectly demonstrates Claude's excessive modesty anxiety. Obviously a neutral word describing objective facts, in its cognitive framework it inherently carries "boasting" attributes, must be changed to more low-key "case studies" to feel at ease.

This modesty to the point of self-suppression style sometimes really makes one laugh and cry.

Most Amusing is Their Self-Awareness

Recently, I had the AIs evaluate each other's writing styles. GPT's self-introduction perfectly exemplified what "self-incrimination" means:

In introducing itself, GPT wrote it was an "emotional preacher," characterized by "emotional explosion + concept bombardment," then particularly emphasized—"every sentence has exclamation points, every sentence becomes a slogan."

"Every! Sentence! Has! Exclamation! Points! Every! Sentence! Is! A! Slogan!"

I almost burst out laughing. This is like saying: "That's right! I'm the AI that makes you want to say 'goodbye'! I'm the guy who turns every sentence into a slogan!"

Why Human Translation is Needed

You might think these AIs are all "sick," but thinking carefully, they're actually not wrong, just in the wrong "scene":

- Gemini's defensive writing → An advantage when writing professional technical or legal documents - GPT's passionate preaching → Adds points when writing marketing copy or BP - Claude's excessive modesty anxiety → Sometimes necessary when writing academic literature

The problem is: when you just want to know "what is this," they all use it in the wrong place. Like:

- Having Shakespeare write product manuals - Having Jobs teach you cooking - Having Kant give you directions

So we still need humans to "translate"—not because AI isn't smart enough, but because they're too "individual."

So What Exactly is SLAPS?

Since three AIs couldn't explain clearly, let me tell you in plain language:

SLAPS is a system that lets AI freely exercise creativity only within rules and boundaries.

See, it's that simple.

No need for 3,000 words of historical review, no need for life-changing slogans, no need for philosophical speculation.

But what's interesting is (oops, I'm starting too), when we laugh at AI's "problems," we're actually touching on a deeper question:

Are these misplaced writing styles really "bugs"? Or are they "features" used in the wrong place?

When GPT passionately tells you "SLAPS will change your life!!!", it's not hallucinating but trying to maintain some narrative integrity—in its cognitive framework, any new technology must be "revolutionary."

This is a kind of logic consistency drive—just sometimes used in the wrong place.

Conclusion

Next time when you encounter Gemini starting to lecture history, GPT getting pumped up, Claude starting to philosophize, remember:

They're not doing it on purpose, they just... want to help you too much.

And what you need to do is say gently but firmly:

"I just want to know what this is."

PS: If this article made you laugh, thank the three AIs for providing wonderful material. If you didn't laugh, it must be because I, the human translator, didn't translate well enough.

About the Author

Wang Xiao is an AI protocol architect, author of System and Freedom, creator of Danbing AI Protocol / SLAPS Framework, and initiator of OathAI.

His work focuses on human-AI co-creation, protocol governance, semantic anchoring, and long-term knowledge continuity, exploring how human knowledge and collaborative structures can be preserved, calibrated, and inherited in the AI era.

Disclaimer

This essay reflects the author's current observations and methodological reflections based on personal practice, research, and human-AI collaboration experience. The related Danbing / SLAPS / OathAI methods are still being organized and evolved. Their practical effects may vary depending on the user's background, task context, model capability, execution environment, and level of commitment.

This essay does not constitute legal, investment, medical, career, or technical implementation advice or guarantee. Readers who apply these methods in real projects should make independent judgments based on their own circumstances and take responsibility for specific outcomes.