Multilingual AI Detector: Originality.ai 2.0.0 Accurately Detects AI Content in 30 Languages
Originality.ai is thrilled to announce that our AI Detection Multi Language Model Expansion is now available! Confidently detect multilingual AI content in 30 Languages with Originality.ai.
Not only are we expanding our AI detector’s language capabilities, but our model is also showing significant improvements in its multilingual AI model accuracy.
As AI-generated content becomes increasingly prevalent and sophisticated, the ability to accurately detect such content across multiple languages is crucial.
Our Multilingual AI Detector aims to address this challenge by providing a robust tool for identifying AI-generated text across a wide range of languages.
Let’s dive into the latest in multi-language AI detection!
Key Takeaways (TL;DR)
The Originality.ai Multi Language 2.0.0 model now supports 30 languages!
Multi Language 1.3.0 (supported 15 languages): Russian, Spanish, Turkish, Italian, French, German, Portuguese, Dutch, Chinese Traditional, Chinese Simplified, Greek, Polish, Vietnamese, Japanese, and Persian (plus English)
Multi Language 2.0.0 (15 additional languages are now available): Korean, Ukrainian, Indonesian, Hindi, Standard Arabic, Czech, Swedish, Danish, Thai, Urdu, Slovak, Norwegian, Bulgarian, Romanian, and Finnish.
Notable multilingual AI model accuracy improvements with an overall accuracy of 97.8%
Reduced false negatives to 1.99% and lowered false positive rate to 2.4%
Quick Model Comparison: Lite, Turbo, and Multi Language
Our Lite and Turbo models (for English language detection) still outperform the current multilingual model. As a quick refresher:
Lite (allows light AI editing like Grammarly): 98% accuracy and an under 1% false positive rate
Turbo (zero tolerance policy on AI use): 99% accuracy and under 3% false positive rate
Multi Language (30 languages): 97.8% accuracy and lowered false positive rate to 2.4%.
What Languages Does the Originality.ai AI Detector Support?
The Originality.ai Multi Language 2.0.0 AI detection model has been trained on 30 languages from our benchmark dataset.
Originality.ai Multi Language 2.0.0 now supports AI detection in the following languages:
To scan for AI content in one of these 30 languages, it’s as easy as signing up, and navigating to the Content Scanner. Then, selecting AI Check and Multi Language. Simple.
Alternatively, you can continue to use our Lite and Turbo models for detecting AI Content in the English language.
We evaluated our Multi Language AI Detection Model, both the previous version and Multi Language 2.0.0, on the same benchmark dataset containing 127,150 samples across 30 languages.
The dataset included:
Human-written content (63,575 samples)
AI-generated content (63,575 samples) from 12 different AI models
Generative AI is always evolving with new releases and updated capabilities. So, for our testing, we included content samples from 12 AI models, including some of the latest and most popular AI models such as:
Claude 3.5 Haiku
Claude 3.7 Sonnet
GPT-4o
GPT-4o mini
Gemini 2.0 Flash
Evaluation Metrics
We evaluated both models using standard classification metrics:
Accuracy: Percentage of correctly classified samples
Precision: Percentage of samples identified as AI-generated that are actually AI-generated
Recall: Percentage of actual AI-generated samples that are correctly identified
F1 Score: Harmonic mean of precision and recall
False Positive Rate (FPR): Percentage of human-written content incorrectly classified as AI-generated
False Negative Rate (FNR): Percentage of AI-generated content incorrectly classified as human-written
Overall Performance
Let’s take a look at the overall performance of Originality.ai Multi Language 2.0.0.
Metric
Originality.ai Multi Language 2.0.0
Accuracy
97.81%
Precision
97.61%
Recall
98.01%
F1 Score
97.81%
False Positive Rate (FPR)
2.40%
False Negative Rate (FNR)
1.99%
Multi Language 2.0.0 Performance Highlights
Both human-written and AI-generated content detection improved in v2.0.0:
Human-Written Content: Accuracy increased to 97.6%
AI-Generated Content: Accuracy increased to 98.01%
False Positive Rate: Reduced to 2.40%
False Negative Rate: Reduced to 1.99%
Accuracy by Language
The Originality.ai Multi Language 2.0.0 AI Detector exhibits consistent cross-language performance and achieves high accuracy across all languages, with minimal variation between languages.
Check out the table and graph below to review accuracy metrics by language:
Language
Accuracy
Precision
Recall
F1
FPR
FNR
th – Thai
98.65%
98.92%
98.45%
98.68%
1.14%
1.55%
ur – Urdu
99.48%
99.64%
99.35%
99.50%
0.37%
0.65%
cs – Czech
97.79%
97.84%
97.84%
97.84%
2.27%
2.16%
fi – Finnish
97.59%
97.36%
97.92%
97.64%
2.75%
2.08%
sk – Slovak
98.14%
97.90%
98.45%
98.18%
2.18%
1.55%
uk – Ukrainian
98.41%
98.88%
97.91%
98.39%
1.10%
2.09%
el – Greek
98.34%
98.54%
98.19%
98.36%
1.51%
1.81%
zh – Chinese*
98.76%
99.63%
97.93%
98.77%
0.38%
2.07%
pl – Polish
98.25%
97.81%
98.60%
98.20%
2.07%
1.40%
vi – Vietnamese
98.43%
98.51%
98.39%
98.45%
1.53%
1.61%
bg – Bulgarian
98.42%
98.33%
98.60%
98.46%
1.77%
1.40%
ar – Arabic
99.30%
99.26%
99.34%
99.30%
0.73%
0.66%
sv – Swedish
98.42%
98.65%
98.23%
98.44%
1.38%
1.77%
no – Norwegian
97.85%
97.52%
98.31%
97.91%
2.62%
1.69%
nl – Dutch
97.95%
97.80%
98.26%
98.03%
2.37%
1.74%
fa – Persian (Farsi)
98.91%
98.86%
99.00%
98.93%
1.18%
1.00%
ro – Romanian
98.73%
99.31%
98.10%
98.70%
0.66%
1.90%
ru – Russian
98.38%
98.42%
98.19%
98.30%
1.44%
1.81%
it – Italian
98.62%
98.77%
98.45%
98.61%
1.21%
1.55%
da – Danish
98.24%
98.11%
98.40%
98.25%
1.92%
1.60%
de – German
98.36%
98.24%
98.54%
98.39%
1.82%
1.46%
ko – Korean
98.59%
99.44%
97.74%
98.58%
0.55%
2.26%
hi – Hindi
99.44%
99.77%
99.07%
99.42%
0.21%
0.93%
tr – Turkish
98.38%
98.57%
98.10%
98.33%
1.35%
1.90%
ja – Japanese
98.22%
98.72%
97.72%
98.22%
1.27%
2.28%
fr – French
98.95%
99.30%
98.62%
98.96%
0.70%
1.38%
pt – Portuguese
98.62%
98.78%
98.30%
98.54%
1.10%
1.70%
id – Indonesian
99.40%
99.65%
99.17%
99.41%
0.36%
0.83%
es – Spanish
98.40%
98.23%
98.04%
98.13%
1.33%
1.96%
*Note: zh – Chinese language metrics include Chinese Traditional and Chinese Simplified.
Final Thoughts & What’s Next
We’re very proud of the significant improvements our team has made with Multi Language 2.0.0, but we’re not done yet!
Our focus will always be on leading the conversation in AI content detection, its importance, and the value of openness and transparency.
Here’s a look at what’s coming next:
Expanding language support: Adding support for additional languages!
Adaptation to new AI models: Continuously updating the detector to recognize content from emerging AI models.
Continue enhancing robustness to adversarial techniques: As AI use increases, so too do platforms designed to ‘bypass’ AI detection. So, to ensure our AI detection model stays robust, we’ll continue enhancing the detector to identify content which has been deliberately modified to attempt to evade detection.
We hope you enjoy our latest release! If you have any questions, you can always reach out and contact support to chat with our exceptional customer service team.
Founder / CEO of Originality.ai I have been involved in the SEO and Content Marketing world for over a decade. My career started with a portfolio of content sites, recently I sold 2 content marketing agencies and I am the Co-Founder of MotionInvest.com, the leading place to buy and sell content websites. Through these experiences I understand what web publishers need when it comes to verifying content is original. I am not For or Against AI content, I think it has a place in everyones content strategy. However, I believe you as the publisher should be the one making the decision on when to use AI content. Our Originality checking tool has been built with serious web publishers in mind!
