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 doesn't just provide a binary "Yes" or "No." It performs a deep architectural analysis of your text, identifying patterns typical of LLMs while respecting the nuances of academic writing.

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.

Humanizing at LilyGuides is not about "spinning" text to cheat. It's about restoring the natural variance and "burstiness" that AI tools often strip away. Our humanizer acts as a sophisticated writing coach, suggesting ways to inject your personality back into your work.

We focus on authenticity-first humanizing. Instead of generating entirely new content, our models analyze your intent and provide guidance on how to restructure sentences to reflect a genuine human perspective. This process maintains your creative ownership while ensuring your work meets the highest standards of originality.

Our models are built on a foundation of real-world participation. We have trained on an extensive dataset of original student and research work dating from 2009 to the present - all volunteered by users who believe in this mission.

The - training examples represent a massive 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. This dynamic dataset grows with every new model shipment, ensuring our accuracy stays ahead of evolving AI.

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

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