What GPTZero measures, and how to read human on it
GPTZero is the detector most students meet first, because it is the one most teachers know by name. It is known for two measurements it made famous, perplexity and burstiness, and for highlighting individual sentences it believes are machine-written. Humanize 360 measures the same two things, plus nine more, and shows you the full breakdown before you rewrite a word.
One thing first. No tool, ours included, can promise what GPTZero will say about a piece of text. It updates its model without notice, and it disagrees with Turnitin, Originality and Copyleaks on the same paragraph. What you can control is whether your writing carries the statistical tells that all of them look for. That is what our engine changes, and it is the only honest thing anyone in this category can offer.
Perplexity and burstiness in plain words
Perplexity is how surprising the next word is. A language model chooses the likeliest word almost every time, so machine text has low perplexity. Human writers pick odd words, repeat themselves, use slang, and change their minds mid-sentence. Burstiness is how much sentence length and rhythm vary across a passage. People write a long winding sentence and then a short one; models tend not to. According to its public documentation, GPTZero combines these with a trained classifier and reports a document-level verdict alongside sentence-level highlighting.
Why your own essay can light up
- You were taught to write formally: no contractions, no first person, every sentence a similar length. That is exactly the pattern low burstiness describes.
- You are writing in your second or third language and reach for the safest, most common word each time. Safe words are predictable words.
- You followed a template: topic sentence, three supporting points, summarizing line. Every paragraph the same shape reads as structural sameness.
- You wrote about a technical subject with a fixed vocabulary, where there is only one right word and you used it every time.
Writing for GPTZero with Humanize 360
The signals in our detector that matter most here are sentence-length variance, word predictability and structural sameness. Check the text in the box above and look at those three. Then run the humanizer: Academic mode at Balanced strength for coursework, which keeps the formal register and protects citations and numbers; Standard or Flowing for a blog post or personal statement. Check again. If variance is still flat, run Heavy once, or paste 300 words of your own earlier writing as a style profile so the engine matches your natural rhythm. Read the result aloud. If it sounds like you, stop.
If you teach and use GPTZero
A GPTZero verdict is a probability from a model, not evidence of what a student did. Sentence highlighting looks precise, but it is still a guess about style. Compare the submission with the student's earlier work, ask them to explain their argument, and follow your institution's process. Our page for teachers at /ai-detector-for-teachers shows the eleven signals so a conversation can start from something concrete, and /ethics sets out where we stand on academic use: we do not support using any tool to breach an institution's rules.
Common questions
Does Humanize 360 make text read human on GPTZero?
It removes the statistical tells GPTZero is known for measuring: flat sentence rhythm, predictable word choice, uniform paragraphs. We cannot promise a verdict, because GPTZero changes its model and no tool controls another tool's output.
Which mode should I use for an essay?
Academic mode at Balanced strength. It keeps a formal register, leaves citations and figures untouched, and still varies sentence length. Use protected words for names and technical terms you cannot change.
Why did GPTZero flag something I wrote myself?
Usually low burstiness. Very even, careful, formal writing shares that trait with machine text, especially from non-native writers. Our detector shows which signal is responsible so you can see it is register rather than stock phrases and tell-words.
Is my text sent to GPTZero or any other company?
No. Our detector and humanizer are our own code on our own servers. Anonymous runs are processed in memory and discarded. Nothing is sent to GPTZero, OpenAI or anyone else.