Originality 5.0 and 5.0 Max
Originality 5.0 and 5.0 Max help you review human-written, AI-generated, and mixed human–AI content with clearer results and improved detection accuracy. Both models offer updated AI scores, improved sentence highlighting, and automatic language detection.
Which model should I choose
Model | Best suited for | Speed and Cost |
|---|---|---|
5.0 | Everyday scans and scanning content at scale | Faster results at a lower credit cost |
5.0 Max | Important reviews where the highest available detection accuracy is preferred | Longer scan times and a higher credit cost |
You can switch between models on the app. Both models support Multilingual writing.
Understanding your AI score
The updated results show the percentage of your text classified as:
- Likely Original (Under 10%): Text the model identifies as likely human-written.
- Mixed writing (Between 10%-80%): Text the model identifies as involving a combination of human and AI writing.
- Likely AI (Between 81-100%): Text the model identifies as likely involving AI.
This differs from the confidence scores used by previous Originality models. The new scores describe the model’s assessment of AI involvement in the text, rather than its confidence that AI use exceeds a selected allowance. Because the scoring approach has changed, avoid treating an older confidence score and a new 5.0 score as equivalent measurements.
AI detection can make mistakes. Treat results as a signal for further review, rather than proof of authorship or misconduct.
AI Score Setting
AI Score Setting helps you interpret the score based on how much AI assistance you allow, whether that's a little AI help or more substanial assitance. Your AI score won't change when you adjust the setting.
Use the slider to select from these preset levels and see how your result is classified based on your expectations for AI use:
- 5%: Minimal AI - For writing where you expect very little AI assistance, such as a personal essay with occasional AI-assisted spelling or gramma corrections.
- 10%: Light AI (recommeneded) - For writing where limited AI support is acceptable, such as using AI to polish a few sentences or improve clarity while keeping the writing your own.
- 30%: Moderate AI - For writing where more AI involvement and hybrid human-AI writing is acceptable, such as using AI to rewrite several passages or expand on ideas you developed.
- 65%: Substantial AI - For writing where extensive AI support is acceptable, such as starting with an AI-generated draft and then reviewing, editing, and adding your own insights.
* These examples illustrate possible expectations for AI use; they do not mean that a particular editing action will produce a specific score.
Improved sentence highlighting
Updated sentence highlighting helps you review the portions of your document identified as involving AI. Use it alongside the overall result to decide which passages need a closer look.
Multilingual detection
Both 5.0 and 5.0 Max automatically identify the language of your text, so you no longer need to select the separate Multi Language model.
The models support AI detection across more than 30 languages. Detection performance can vary by language and type of content.
Accuracy and limitations
5.0 and 5.0 Max are our most accurate AI detection models yet, with improvements in detecting mixed human–AI writing as well as fully human-written and AI-generated content.
In our evaluation of 1,000 mixed documents from OpAI-Bench, 5.0 classified 82.9% as mixed writing, and 5.0 Max classified 90.1% as mixed writing. These results describe performance on that test dataset and do not guarantee the same results for every document.
The models also support code detection. However, testing on code showed lower performance than on long-form text, so code results require particular care when interpreting them.
For more information, read the blog.
Where can I use 5.0 and 5.0 Max
5.0 and 5.0 Max are available through:
- The Originality.ai app
- Bulk Scan
- Chrome and Firefox browser extensions
- API
- MCP
- Moodle
Updated on: 05/11/2026
Thank you!