← aismells

Field Guide

Signs of AI Writing

AI writing tells are repeated patterns that make text sound like ChatGPT: familiar phrases, inflated word choice, mechanical sentence shapes, generic enthusiasm, and formatting habits that appear too often to feel intentional.

Check text with Sniff Test

The Short Version

The tell is density, not presence. A human can write "it is worth noting" or use an em dash. The signal gets stronger when the same draft also has "delve," three-item lists, rhetorical questions, padded transitions, uniform paragraphs, and bold-first bullets.

Use these patterns as an editorial filter, not as proof of authorship.

Common AI Writing Tell Categories

Overused AI Words

Words like "delve," "leverage," "robust," "tapestry," and "nuanced" can cluster in AI-generated text.

See vocabulary tics

ChatGPT Sentence Patterns

Negation reframes, self-posed questions, and rule-of-three padding make prose feel templated.

See sentence tells

Assistant Tone

"Let's unpack this," "here's the thing," and world-historical stakes are clues that the assistant voice leaked through.

See tone tells

Formatting Habits

Excessive bolding, emoji bullets, and dense em-dashes are easy to scan before reading closely.

See formatting tells

How To Tell If Something Was Written By AI

  • Scan for obvious ChatGPT phrases and overused AI words.
  • Look for repeated sentence templates across paragraphs, not one-off examples.
  • Check whether every paragraph has the same length, shape, and transition style.
  • Watch for generic caution, generic enthusiasm, and generic importance.
  • Use a pattern checker, then review the text manually before making any claim.

Paste The Draft

The Sniff Test checks text locally against the aismells catalog. It does not claim certainty. It shows which patterns appeared, where they appeared, and how dense they are.

Run the AI writing detector