GPT Detection
Detect text generated by GPT-4, GPT-3.5, and earlier GPT models
Free AI writing check
Free AI Text Detector
Paste writing to check for patterns associated with ChatGPT, Claude, Gemini, and other language models. Get a free probability score before creating an account.
What We Detect
Comprehensive LLM-generated text detection
Claude Detection
Identify Claude-generated text across different versions and styles
Gemini Detection
Detect Google Gemini and Bard-generated content
LLM Patterns
Identify common LLM writing patterns and stylistic markers
Mixed Content
Detect human-AI hybrid text with confidence estimation per segment
Multilingual
Language-aware review with performance that can vary by language, domain, and text length
Explainable Evidence
Understand exactly why text was flagged
Evidence Types
- Perplexity and burstiness scores
- Stylometric signal analysis
- Sentence-level confidence bands
- Linguistic pattern detection
Operational Guidance
- Processing time depends on text length and scan mode
- Batch workflows are available for document collections
- Short and heavily edited text requires cautious interpretation
Limitations & Best Practices
Understanding model constraints for optimal results
False Positive Risk
Professional writing, edited text, or short samples may increase false positives.
Adversarial Attacks
Human paraphrasing and AI obfuscation tools may evade detection.
Generalization Bounds
Performance varies by domain and writing style. Best on academic/professional content.
API Integration
Get started in minutes with our unified API
curl -X POST https://app.zerotrue.app/api/v1/check \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"input": {
"type": "text",
"value": "Sample text content"
},
"is_private_scan": true,
"is_deep_scan": false,
"idempotency_key": "req-123"
}'Frequently Asked Questions
Which LLMs can you detect?
The detector evaluates signals associated with major LLM families, but coverage and confidence vary across model versions and editing.
How accurate is text detection?
Accuracy varies with text length, language, model family, paraphrasing, and human editing. Use confidence bands and review limitations.
Can you detect AI-edited text?
Human editing can weaken or change AI-associated signals. Mixed content may still produce indicators, but it needs cautious review.
Does it work on all languages?
Performance varies by language, domain, length, and model family. Test the languages required by your workflow.
Is my text data stored?
Retention can depend on the plan and scan mode. Confirm current retention and deletion requirements before submitting sensitive text.