Perplexity and Burstiness Checker

Measure perplexity and burstiness directly, the two signals behind nearly every AI detector. Free, per-sentence, explained.

0 words

Tested against

  • Turnitin
  • GPTZero
  • Originality.ai
  • Copyleaks
  • ZeroGPT
  • Winston AI
  • Scribbr
  • Writer
  • Sapling
  • Crossplag
  • Content at Scale

Perplexity is how surprising each word is given the words before it. Model output is low-perplexity by construction, models select likely continuations, so the prose is more predictable than human writing. Burstiness is the variance in sentence length: people write a fourteen-word sentence, then a four-word one, then a thirty-word one. Models tend not to.

Seeing both per sentence is more actionable than any composite score. Low burstiness is fixed by varying sentence length. Low perplexity is fixed by specific, unexpected word choices, which usually means adding concrete detail rather than reaching for a thesaurus.

FAQ

It is the same concept, word-level predictability, estimated from measurable text features rather than by running a model. The direction and relative ranking are what matter.

Higher is more human. There is no target number; what matters is that your sentence lengths genuinely vary rather than clustering around one value.

You can, and it reads terribly. Specific concrete detail raises it in a way a thesaurus never will.

The writing analysis and your Style DNA run in your browser. For a real AI score, the text you check is sent to an AI-detection provider, and we never store it. When you rewrite on a paid plan, the draft is sent to our server, rewritten and sent back without being stored. AI features (rewriting, humanizing and Copilot) send the text to Cloudflare Workers AI, which processes it and doesn't keep it.

Explore more

All free tools
Sound like you, not AIFree AI scan · rewrites in your voiceGet started