Safeguarding Human Authenticity

"LilyGuides is a research project dedicated to helping students and scholars maintain authenticity. We guide you on specific areas to rectify in your work, fostering creativity rather than providing ready-made shortcuts that hinder academic growth."

LilyGuides was born out of a commitment to public participation and academic excellence. Originally sustained by community funding, our core value remains unchanged: ensuring that your unique voice is heard, respected, and accurately represented.

The rapid emergence of AI has created an unintended crisis in academia. Scholars and students are increasingly struggling to prove the authenticity of their original work as AI detectors become more aggressive and less reliable.

We've observed a growing trend where detectors frequently misclassify human-written material as machine-generated. This "AI Trap" doesn't just threaten grades; it threatens the very foundation of research and creative writing.

Our AI Detection functionality is highly effective, achieving over 95% accuracy with a false positive rate of just 1.3%. This precision allows us to correctly categorize user text and phrases as either human-written, AI-generated, or mixed (AI-humanized content).

By highlighting specific sentences and scoring perplexity and burstiness, we empower you to understand why a piece of writing might be flagged. This transparency allows you to refine your work and maintain your unique scholarly voice without fear of false accusations. All our AI detection models are entirely free to use, with no payment required.

We also offer the most accurate humanizer model available, intended specifically to guide students and researchers in improving their original work. While our humanizer is fully capable of converting100% AI-generated content into a 100% human-written version, its primary purpose is to act as a sophisticated writing coach to restore natural variance.

Instead of generating entirely new content for you to submit, our models analyze your intent and provide guidance on how to restructure sentences to reflect a genuine human perspective. This ensures your work meets the highest standards of originality. Like our detectors, the humanizer models are completely free for all users.

Our exceptional accuracy has been achieved through training our models using over 20,000,000 datasets of real human-written research papers and essays from authors and students who have contributed to the project. We utilized papers spanning from 2009 to 2017 as the original human benchmark, and generated the AI versions of those same papers to train our systems.

These massive datasets represent a collaborative effort. Unlike generic models trained on scraped web data, LilyGuides uses curated, verified human-written material to ensure that our detectors can distinguish between complex human thought and repetitive AI patterns.

Check our Transparency Page for detailed model performance and training metrics.

Yes, our humanizer bypasses all major tested AI detectors. However, its usage must strictly align with our Terms and Conditions.

The intended use of the humanizer is to produce structural templates that guide you in restructuring your work to avoid unfair penalization by overly aggressive detectors.

We strictly prohibit users from directly using the humanized text as their own original work. Failure to adhere to this policy will result in a permanent ban from our platform.

The mission of LilyGuides is supported by global research highlighting the limitations and biases of current AI detection technologies.