Skip to content

Instantly share code, notes, and snippets.

@leegonzales
Created December 19, 2025 04:37
Show Gist options
  • Select an option

  • Save leegonzales/c968df445a9ed95f2b93bb3b67817341 to your computer and use it in GitHub Desktop.

Select an option

Save leegonzales/c968df445a9ed95f2b93bb3b67817341 to your computer and use it in GitHub Desktop.
Prose Polish: Four Writing Samples Compared

Prose Polish Comparison: Four Writing Samples

Score Summary

Sample Score Verdict Key Patterns
AI-Generated 89/100 Almost Certainly AI "delve," "landscape," false contrarian, no specifics, "In conclusion"
CNN News 44/100 Probably AI / Genre-Formulaic Formulaic structure, good specificity, both-sides balance
Wikipedia/Britannica 32/100 Possibly AI-Assisted (genre-appropriate) Dense academic style, high specificity, appropriate hedging
Economist ~18/100 Likely Human Wit, concrete analogy, commitment, no throat-clearing

What the Scores Mean

89 (AI-Generated): Every AI pattern present. No specificity, no commitment, no voice. Template structure with hedging throughout. Rhetorical question ending. This is what unedited ChatGPT output looks like.

44 (CNN): Formulaic news writing. High specificity saves it (real numbers, named sources, dates). But structural patterns match AI: inverted pyramid, both-sides balance, no distinctive voice. Could be human journalism or AI mimicking news style.

32 (Wikipedia/Britannica): Academic encyclopedia style. Hedging appropriate for historical uncertainty. Dense information with specific dates, names, places. Score reflects formal patterns that are genre-appropriate, not AI tells.

~18 (Economist): Opens with Groucho Marx analogy, not "In today's..." Makes a claim immediately. Wit serves argument. No hedging. This is what committed, voiced prose looks like.


Pattern Comparison

Opening Lines

Sample Opening Score Impact
AI-Generated "In today's fast-paced business environment..." +15 (Corporate Opening Gambit)
CNN "Even if you believe you haven't yet been directly impacted..." +5 (direct address, but soft)
Wikipedia "The Byzantine Empire represented the eastern half..." +0 (direct statement of fact)
Economist "Groucho Marx memorably said..." -5 (unexpected cultural reference)

Specificity

Sample Specific Details Score Impact
AI-Generated None +20
CNN $125K, 14.5M deaths, $12.5T, named sources -15
Wikipedia 330 CE, 1453, Constantine, Bosporus -15
Economist Groucho Marx, eurozone economies -10

Commitment Level

Sample Commitment Score Impact
AI-Generated Zero — all hedged +15
CNN Low — reporting not arguing +5
Wikipedia Medium — claims with appropriate uncertainty +3
Economist High — "Some of Europe's economies may feel..." -5

Lessons

  1. Specificity matters most. Real names, numbers, dates dramatically lower scores.

  2. Genre affects interpretation. A 44 for news journalism is different than a 44 for an essay. Some patterns are conventions, not tells.

  3. Opening lines are diagnostic. "In today's..." is an 85%+ AI indicator. Unexpected cultural references suggest human.

  4. Commitment separates. AI hedges everything. Humans take positions.

  5. The score is a diagnostic, not a judgment. A 32 for Wikipedia is fine—that's what encyclopedias sound like. An 89 for thought leadership is not.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment