AI Detector

Definition¶

An AI detector is a software tool that examines a piece of text and estimates the probability that it was produced by an AI language model rather than written by a human. Most detectors output a percentage score ("85% likely AI-generated") alongside a sentence-level breakdown showing which parts triggered the detection signals.

How AI Detectors Work¶

The most common detection methods measure two statistical properties:

• Perplexity — how predictable each word choice is given its context. AI-generated text tends to have lower perplexity because language models select statistically likely tokens.
• Burstiness — how much sentence length varies throughout a document. Human writing varies more; AI writing is more uniform.

More advanced detectors also use trained classifiers that learn to recognize stylistic patterns in AI output — including stock phrases, passive construction frequency, and hedging language density.

Popular AI Detectors¶

• GPTZero — built for academic use, shows sentence-level probability
• Originality.ai — marketed to content publishers, includes plagiarism checking
• Turnitin AI Detection — integrated into the widely-used academic submission platform
• Copyleaks — covers AI detection alongside plagiarism checking
• Winston AI — focuses on education and publishing use cases

Limitations¶

AI detectors are probabilistic, not definitive. Documented limitations include:

• False positives — dense academic writing, ESL writing, and consistent formal styles are sometimes flagged as AI-generated even when written by humans
• Evasion — humanized text (text that has been rewritten to reduce low perplexity and increase burstiness) can fall below detection thresholds
• Model dependency — detectors trained on GPT output may not detect text from other models as reliably

Related Terms¶

Perplexity · Burstiness · AI Humanizer · False Positive Rate