Why Do LLMs Like to ‘Underscore’ Everything: Common AI Jargon and How to Humanize it

AI has a common way of writing, and always “underscores the pivotal importance” of everything and anything.

In 2026, academic writing at UM, USM, and UTM is suffering from a specific linguistic infection. It isn’t bad grammar or spelling errors. It is the “AI Jargon Virus.”

When you prompt ChatGPT, Gemini, Claude or Perplexity to “rewrite this paragraph to sound academic,” it doesn’t actually sound academic. It sounds like a press release. It sounds inflated, dramatic, and weirdly obsessed with words like realm, lens, pivotal, and, the worst offender of all, underscore.

To an AI, this language feels “sophisticated.” To a human examiner, it feels like you are trying to hide a weak argument behind big words.

In this guide, we are going to dissect the “Underscore Epidemic.” We will explain why AI models are addicted to these specific words, why they ruin your scholarly authority, and provide a definitive “Translation Dictionary” to turn that robotic fluff back into clean, precise research.

1. The Psychology of the Bot: Why “Underscore”?

Why does an advanced AI model, trained on the entire internet, default to the same five words over and over again?

The “Safety” of Inflation

AI models are prediction engines. Their main goal is to generate the most “statistically probable” next word. When an AI detects that you want a “formal” or “academic” tone, it looks for words that have a high probability of appearing in formal texts.

However, it lacks nuance.

• Human Nuance: We know that underscore is a strong word. We use it maybe once in a whole chapter to highlight a major breakthrough.
• AI Logic: Underscore = Important. Therefore, if I want to show that this sentence is important, I will use underscore.

The result is a thesis where everything is underscored. The methodology underscores the theory. The theory underscores the results. The results underscore the conclusion.

When everything is emphasized, then nothing is emphasized!

The “Void” of Opinion

AI cannot have an opinion. It cannot say, “This result is wrong” or “This theory is garbage.” It has to be polite and neutral.

Words like pivotal, crucial, and essential are “Empty Intensifiers.” They allow the AI to sound supportive and positive without actually saying anything specific.

• Human: “The data shows a 5% drop.” (Specific).
• AI: “The data underscores a pivotal shift in the landscape.” (Vague, dramatic, safe).

2. The “Dirty Dozen”: AI Words That Trigger Examiners

If you want to know if a student used AI, just Ctrl+F for these words. In my editing work at Proofreading Malaysia, I call these the “Dirty Dozen.” If they appear frequently, the text is almost certainly generated.

1. Underscore / Underscoring

• The AI Obsession: It loves to use this to connect a finding to a conclusion.
• Why it fails: It’s passive. It sounds like a corporate memo.
• The Fix: Demonstrates, shows, indicates, highlights.

2. Pivotal

• The AI Obsession: Everything is “pivotal.” A small sample size limitation? Pivotal. A major discovery? Pivotal.
• Why it fails: In real research, very few things are truly “pivotal” (meaning they turn the entire course of the field).
• The Fix: Important, significant, key, central.

3. Lens

• The AI Obsession: “Through the lens of…”
• Why it fails: It’s a cliché metaphor. Academic writing should be direct, not poetic.
• The Fix: From the perspective of, using the framework of, based on.

4. Realm

• The AI Obsession: “In the realm of biology…” or “In the digital realm…”
• Why it fails: It sounds like Game of Thrones. We are researchers, not wizards. We work in “fields,” “disciplines,” or “contexts,” not realms.
• The Fix: Field, domain, area, context.

5. Delve

• The AI Obsession: “This chapter delves into…”
• Why it fails: This is the #1 tell-tale sign of ChatGPT. Humans rarely use “delve” in 2026. We “examine,” “investigate,” or “explore.”
• The Fix: Investigate, examine, analyze, discuss.

6. Tapestry

• The AI Obsession: “A rich tapestry of culture…”
• Why it fails: Again, too poetic. Unless you are studying actual textiles, do not use this word.
• The Fix: Complex mix, variety, combination.

7. Landscape

• The AI Obsession: “The ever-evolving landscape of cybersecurity…”
• Why it fails: It’s a filler phrase. It adds zero information.
• The Fix: Current state, field, context.

8. Testament

• The AI Obsession: “This result is a testament to the resilience…”
• Why it fails: Too dramatic. It sounds like a eulogy.
• The Fix: Proof, evidence, indication.

9. Foster

• The AI Obsession: “To foster a sense of community…”
• Why it fails: It’s corporate HR speak. In research, we “promote,” “encourage,” or “facilitate.”
• The Fix: Encourage, promote, enable.

10. Nuanced

• The AI Obsession: “A nuanced understanding…”
• Why it fails: AI uses this when it doesn’t actually understand the details. It’s a way of saying “complicated” without explaining how.
• The Fix: Detailed, complex, specific.

3. The “Inflated Sentence” Syndrome

The problem isn’t just individual words. It’s how AI combines them into “Inflated Sentences.”

An AI sentence is like a balloon—it looks big, but it’s mostly air.

The AI Sentence:

“This study underscores the pivotal importance of adopting a nuanced approach to navigate the complex landscape of urban planning in the digital realm.”

The Human Translation:

“This study shows that urban planning in digital contexts requires a specific, detailed approach.”

Do you see the difference?

• The AI version is 24 words. It uses “underscore,” “pivotal,” “nuanced,” “landscape,” and “realm.” It is exhausting to read.
• The Human version is 14 words. It uses zero jargon. It is clear, direct, and also just straight to the point.

Why Examiners Hate This

When a supervisor reads the AI version, their brain has to work harder to extract the meaning. They have to mentally delete the “fluff” to get to the point.

• They don’t think, “Wow, this student has a great vocabulary.”
• They think, “Why is this student wasting my time?”

4. The “De-Botting” Translation Table

Print this out. Tape it to your monitor. When you see the word on the left in your thesis, replace it with the word on the right.

The “Significance” Inflaters

The “Metaphor” Traps

The “Action” Fakers

5. The “Before & After” Case Study

Let’s look at a real example of an abstract I edited recently.

The AI Draft:

“This study delves into the realm of renewable energy in Malaysia. It underscores the pivotal role of solar panels in fostering a sustainable energy landscape. The results are a testament to the nuanced challenges faced by policymakers.”

The Human Rewrite:

“This study investigates renewable energy adoption in Malaysia. It demonstrates the critical role of solar panels in promoting sustainable energy infrastructure. The results highlight the specific challenges faced by policymakers.”

The Difference:

The Human version is grounded. It swaps “landscape” for “infrastructure” (specific). It swaps “nuanced” for “specific” (measurable). It sounds like an expert wrote it, not a PR machine.

6. The “Tone Check” Protocol

Before you submit your thesis, you need to perform a “Tone Check.” This is different from a grammar check. You are checking for Voice.

1. The “Read Aloud” Test: Read your Abstract out loud. If you run out of breath in the middle of a sentence because it has too many “underscores” and “moreovers,” it’s AI. Humans write with natural pauses.
2. The “Specifics” Scan: Search for the word nuanced. If you find it, read the sentence. Does the sentence actually explain what the nuance is?
   • Bad: “The situation is nuanced.” (Empty).
   • Good: “The situation is complex because of factor X and factor Y.” (Real).
3. The “Adjective Diet”: Go through your Introduction and delete 50% of the adjectives. AI uses adjectives (pivotal, crucial, comprehensive) to hide a lack of data. Humans use data.

Conclusion: Complexity is not Intelligence

The biggest lie LLMs tell us is that “big words” equal “smart writing.”

In academia, clarity is the only intelligence that matters. Your examiner doesn’t want to know about the “ever-evolving landscape.” They want to know what you found, why it matters, and how you proved it.

If you can strip away the “pivotal realms” and the “underscoring tapestries,” you will reveal the one thing the AI can never generate: Your original contribution to knowledge.

Tired of fighting AI fluff alone? Our AI Humanizing Services don’t just swap synonyms. We surgically remove the robotic tone and restore your concise and to-the-point academic writing.