GPTZero AI Detector API
Product · API / developer platform by GPTZero
It provides document-level and sentence-level insights for text from ChatGPT, Gemini, and Claude with 99.9% accuracy.
- Languages
- Python, Node.JS, C#, Java, PHP
Category: Natural Language Processing
- Platforms
- API
- Limits
- Custom plans available for over 5 million words per month
- Languages
- English, Spanish, French, German, Portuguese, and others
- Accuracy (Mixed Documents)
- 96.5%
- False Positive Rate
- Under 1%
- ESL False Positive Rate
- 1.1% on TOEFL essays
- 96.5% accuracy for mixed documents
- Supports 17+ programming languages
- Sentence-level AI highlights via API
- Detects GPT-6, Gemini, Llama, and Claude
- Zero data retention for API calls
Overview
It is designed for businesses, educational institutions, and developers who need to verify content authenticity at scale within their own applications. The API identifies text generated by major large language models including ChatGPT (up to GPT-6), Gemini, Llama, and Claude. It delivers granular results, including document-level classifications, confidence categories, and sentence-by-sentence probability highlights. Organizations such as Microsoft, Nextcloud, and K16 Solutions use the API to monitor AI-driven misinformation and power analytics. The service supports 17 programming languages and offers scalable plans for high-volume scanning requirements.
Detection Insights
- Document Classification: The API returns a document_classification field that labels text as HUMAN_ONLY, MIXED, or AI_ONLY.
- Confidence Categories: Each classification includes a confidence_category of high, medium, or low, with error rates below 1% when confidence is high.
- Sentence-Level Highlights: The highlight_sentence_for_ai field provides sentence-level highlights to show exactly which parts of the text contributed to the classification.
- Probability Scores: The API provides class_probabilities for each classification to give granular insights into the origin of the text.
Developer Features
- Multi-Language Support: Ready-to-use code is available in 17 different languages, including Node.JS, Python, C#, Java, and PHP.
- Scalable Infrastructure: The API is designed for high-volume scanning, with custom plans available for organizations requiring more than 5 million words per month.
- Data Privacy: The service does not store or collect the documents passed into any calls to the API.
- Production-Grade Reliability: The API uses the same deep learning models that serve over 20 million users on the core GPTZero platform.
Use cases
- Educational platforms integrate the API to provide real-time AI detection and writing feedback within learning management systems.
- AI companies use the API as a data labeler to clean synthetic content from their human training datasets.
- Publishers and newsrooms vet new sources by scanning articles for AI-generated content to maintain editorial integrity.
- Security organizations monitor social media and web feeds to identify and track AI-driven misinformation campaigns.
Specifications
- APIs and SDKs
- REST API with support for 17 languages including Node.JS, Python, C#, Java, and PHP
- Platforms
- API
- Limits
- Custom plans available for over 5 million words per month
- Integrations
- Canvas, Google Classroom, Zapier, Nextcloud, Microsoft Threat Intelligence Center
- Languages
- English, Spanish, French, German, Portuguese, and others
- Accuracy (Mixed Documents)
- 96.5%
- False Positive Rate
- Under 1%
- ESL False Positive Rate
- 1.1% on TOEFL essays
- Supported Programming Languages
- 17
- Max Words per Month (Standard)
- 5,000,000+
- Classification Labels
- HUMAN_ONLY, MIXED, AI_ONLY
- Confidence Categories
- High, Medium, Low
- Data Retention
- None for API calls
How it works
- Get API Key: Developers can sign up on the GPTZero platform to obtain a unique x-api-key for authentication.
- Integrate with Platform: Integrate the REST API into your application using provided code samples in 17 languages like Python or Node.JS.
- Scan Documents: Send POST requests to the predict/text endpoint with the document content and version specified in the JSON body.
- Analyze Results: Receive and parse JSON responses containing document-level classifications, confidence scores, and sentence-level highlights.
FAQ
How do I interpret the results from the API?
The API returns a document_classification field (HUMAN_ONLY, MIXED, or AI_ONLY), class_probabilities, and a confidence_category (high, medium, or low). It also provides sentence-level highlights.
Are you storing data from API calls?
GPTZero does not store or collect the documents passed into any calls to the API. Only aggregate inputs for dashboard users are stored to improve the service.
Can the API detect text from the latest AI models?
The API is continually updated to recognize text from the newest models, including ChatGPT (GPT-3 through GPT-6), Google Gemini, LLaMA, and Claude.
Added by AntSpace from public sources (gptzero.me).