Below is a verbatim copy and paste from an Aristotle email connected with ASU+GSV which starts April 12.
Learnings From Thousands of AI Tutoring Sessions With Aristotle
You’ve likely seen a hundred AI tutor pitches this year. We set out to build one that wasn’t just another chatbot.
I’m Shan, one of the co-founders at Aristotle. As part of our development process, we’ve facilitated thousands of hours of voice tutoring conversations, both human and AI. Here are our biggest learnings on why realtime AI tutoring actually shines, and where it struggles.
1. The frontier AI models are trained to answer, not teach.
We struggled for months to get the most expensive and intelligent LLMs to withhold the answer and instead teach socratically. Finally, we stopped prompting our way around it and started training on actual, human-graded teaching conversations. This has been the unlock – no other AI tutor we know of has done this.
2. Students are more willing to voice confusion with an AI tutor
We consistently see that students disclose their confusions faster to an AI tutor than they do to our human tutors. We see a 26% increase in the number of questions our students ask when partnered with an AI tutor versus a human tutor.
🧠 We’re offering free tutoring access to families in the EdTech Insiders community.
🤝 And, we’re building an Advisory Council of teachers, researchers, administrators, and founders. Please consider joining us. Sign up here
shan@heyaristotle.com – I’d love to connect!
See you at ASU+GSV,
Shan Reddy, Aristotle
© 2026 Alex Sarlin and Ben Kornell
Greenville, South Carolina