I checked Wikipedia’s “Signs of AI writing” page. Here are the AI-writing signs it lists, in English, summarized cleanly.
Signs of AI writing
1) Undue emphasis on significance, legacy, and broader trends
Common phrases/ideas:
- “stands/serves as”
- “is a testament/reminder”
- “a crucial/pivotal/vital/significant/key role/moment”
- “underscores/highlights its importance/significance”
- “reflects broader...”
- “symbolizing its ongoing/enduring...”
Typical pattern:
- The text keeps insisting that something is historically important, transformative, or part of a bigger movement, even when the details are thin.
2) Generic praise or inflated importance
Examples of the kind of language to watch for:
- “revolutionary”
- “iconic”
- “groundbreaking”
- “trailblazing”
- “game-changing”
Typical pattern:
- Specific facts get replaced by broad admiration and grand labels.
3) Vague, broad, or overgeneralized statements
Typical pattern:
- The writing sounds polished but stays abstract.
- It avoids concrete, unusual, or verifiable details.
4) Repetition and formulaic structure
Typical pattern:
- Same sentence patterns repeated again and again
- Predictable transitions
- Paragraphs that feel mechanically organized
5) Overly balanced or overly neutral tone
Typical pattern:
- The text avoids making a clear point.
- It sounds cautiously objective but also shallow or evasive.
6) Citation-style quirks and formatting issues
Examples mentioned on the page:
- excessive boldface
- broken markup
- odd citation formatting
- unusual wiki-style formatting habits
7) “Tell” patterns from statistical smoothing
Wikipedia describes this as LLMs tending to:
- replace specific facts with generic ones
- remove nuance
- make text more “average”
- turn unusual details into safe, broad descriptions
In short:
- a highly specific claim becomes a generic compliment
- detailed observation turns into polished but empty prose
8) Detection caveat: not proof
Wikipedia explicitly warns that:
- these are only signs, not proof
- human writing can look the same
- AI detectors are unreliable
- humans are also poor at identifying AI text reliably
Short version
If you want the simplest checklist, Wikipedia’s main AI-writing signals are:
- too much emphasis on importance
- generic praise
- vague abstraction
- formulaic repetition
- overly smooth, balanced writing
- formatting/citation oddities
- loss of specificity and nuance
If you want, I can also turn this into:
1. a checklist you can use to judge any text, or
2. a prompt that pushes writing toward natural human style.
