GPTZero uses deep learning models that keep pace with AI advancements to deliver precise, reliable results that help you understand and interpret the origin of a piece of text. A sentence-by-sentence classifier calculates the probability that a text was created by AI, giving you both document-level and granular insights.

Independent and internal benchmarking show that GPTZero outperforms competitors at detecting mixed documents (where both AI and human writing has been used) with 96.5% accuracy. Our false positive rate is under 1%, making it one of the most reliable detectors available.

Yes. GPTZero is continually updated to recognize text from the newest large language models, including ChatGPT (GPT-3, GPT-4, GPT-5), Google Gemini, LLaMA, Claude, and others. As new models are released, we adapt and update our detection system.

Since 2022, we've focused on reducing bias for ESL (English as a Second Language) writers. By tagging educational data, adding representative datasets, and using text pre-classification, we've reduced the false positive rate on TOEFL essays to just 1.1%.

Many try to bypass AI detectors by paraphrasing or using homoglyph substitutions. Our model's Paraphraser Shield technology is designed to catch these cases. Even if AI content has been altered to look more human-like, GPTZero can detect it.

No AI detector is 100% accurate, and AI itself is changing constantly. GPTZero performs best of longer texts and English prose. We encourage using it as a conversation starter, and not as the final verdict. GPTZero is the only detector specifically de-biased for ESL writers, lowering false positives to 1%.

Our model is trained on millions of documents spanning various domains of writing including creating writing, scientific writing, blogs, news articles, and more. We test our models on a never-before-seen set of human and AI articles from a section of our large-scale dataset, in addition to a smaller set of challenging articles that are outside its training distribution.

When you run text through our API, you'll see a document_classification field that labels it as HUMAN_ONLY, MIXED, and AI_ONLY. Each classification comes with probabilities (via the class_probabilities field) and a confidence_category field, which can be high, medium, or low. When the confidence is "high", error rates are below 1%. The API provides sentence-level highlights (API users can access this highlighting through the highlight_sentence_for_ai field) so you can see exactly which parts of the text contributed most to the classification.

No. We do not store or collect the documents passed into any calls to our API. For dashboard users, only aggregate inputs are stored to improve the service. See our privacy policy for full details.