New AI classifier for indicating AI-written text
ExternalOpenAI's AI Text Classifier was a free tool designed to detect AI-generated text by distinguishing it from human-written content through probability scores. Trained on outputs from multiple AI models, it offered a simple paste-and-check interface for quick assessments. Though innovative, it was discontinued in July 2023 due to low accuracy rates, including only 26% true positives for AI text and notable false positives. It remains a historical benchmark in AI detection efforts, best for casual, supplementary use rather than critical decisions.
Description
OpenAI's AI Text Classifier was a free tool designed to detect AI-generated text by distinguishing it from human-written content through probability scores. Trained on outputs from multiple AI models, it offered a simple paste-and-check interface for quick assessments. Though innovative, it was discontinued in July 2023 due to low accuracy rates, including only 26% true positives for AI text and notable false positives. It remains a historical benchmark in AI detection efforts, best for casual, supplementary use rather than critical decisions.
Key capabilities
- Distinguishing AI-written text from human-written text
- Providing probability scores for text origin classification
Core use cases
- 1.Quick informal checks on text authenticity
- 2.Supplementary tool in multi-detector workflows
- 3.Educational exploration of AI detection limitations
Is New AI classifier for indicating AI-written text Right for You?
Best for
- Casual users for quick, informal text checks
- Users seeking free supplementary AI detection insights
Not ideal for
- Educators or high-stakes decisions due to inaccuracy
- Non-English text, code, or short content analysis
- Reliable detection needs given discontinuation and evasion risks
Standout features
- Free and accessible to anyone
- Intuitive paste-and-check user interface
- Nuanced probability labels (likely AI-written, possibly AI-written, etc.)
User Feedback Highlights
Most Praised
- Free and easy to use
- Intuitive interface for rapid checks
- Nuanced results beyond binary classification
- Useful as a complement to other detectors
Common Complaints
- Low true positive rate (26% for AI text)
- False positives on human text (9%)
- Easily evaded by minor edits to AI text
- Discontinued by OpenAI due to unreliability
- Poor real-world performance vs alternatives