AI for Education
Our responsible AI work starts with expert-written content
Open content changed who could afford to learn. Every book in our library is written and peer-reviewed by subject-matter experts for accuracy and authenticity. Our AI tools build on that vetted content and are designed to keep instructors in control and help students reason, connect, and eventually understand.
Read why open AI matters https://openstax.org/blog/responsible-ai-in-education orange OpenStax_August_2023-53 Stack of OpenStax textbooks in a pile on a table right #f0fef9 3 #0A5B50 12 3 dark-teal-divider 3px 100% centerAI tools
AI designed to keep students thinking
Our tools are designed to make them reason their way there, drawing on our peer-reviewed, openly licensed library, with instructors deciding what reaches their students.
0 80% centerActive learning lessons
EduStax is a prototype AI tool that creates complete active learning lessons for teachers. Learning is constructed by explaining, questioning, and applying ideas, and EduStax generates lessons where students do that work themselves. In a blind study, teachers scored these lessons on par with an expert's and mostly could not tell which were AI-generated. It is now in Algebra 1 classrooms with an efficacy study underway.
#F26B39Guided dialogues
Guided dialogues are AI-led conversations in Assignable that ask students to think critically, reflect on their own understanding, and reason through decisions. Faculty anchor dialogues in specific textbook material and define the questions and structure. The AI deepens a student's responses and clarifies misconceptions, and it is designed not to hand over answers. Available in select Economics activities, with Life Sciences in development.
#002E6DMath assessment practice
Quezzio, powered by Wolfram, brings deeper math assessment to Assignable. It reads the mathematics behind a student's answer, so equivalent forms are accepted, structural errors are flagged, and partial credit can be awarded where the reasoning earns it. Each student receives a unique version of every problem, and when an answer is wrong, the feedback is step-by-step and specific to their solution path. Available in Elementary Algebra 2e.
#FFE665 rounded 3 0 100% 0 #ffffff 2"We’ve seen the global impact of open content and open pedagogy – I believe that the open education movement’s third frontier is open AI tools."
Prof. Richard G. Baraniuk
Founder & Director, OpenStax
Trusted content
AI grounded in OpenStax content
An AI is only as reliable as its source material. When it draws on peer-reviewed, expert-written content, it has less room to invent and a stronger foundation for its answers. Grounding in vetted material is one of the strongest responses to hallucination, and our openly licensed library provides that foundation for any tool, ours and our partners' alike.
Google has brought OpenStax content into its products. Gemini Notebook made the library its first public content partner, and Gemini cites it in responses. Learn Your Way was built on it, with measured gains in a randomized trial.
Microsoft brought OpenStax content into its Learning Zone.
100% 0 0 ai - 2 decorative 2 2 #F3FFF2 #F0FEFE rightEvidence
Whether AI helps a student depends on how it's designed
Students who use general-purpose AI produce higher-quality work but learn less. That gap, documented by the OECD in 2026, narrows when tools are designed around how people actually learn. Learning tends to stick when a student has to reason their way to an answer, and the most useful AI keeps that effort in place instead of removing it.
Read about productive struggle
57% of college students now use AI in their coursework each week (Lumina-Gallup, 2026) green 100% 0 0 ai - 1 decorative 0 2 #F3FFF2 #F5FFEC left ai_research to topAI for Ed RESEARCH
What we're learning about AI
Our current research asks what an AI would need to understand about a learner to teach them well. In a 2026 study, our team measured what happens when AI systems are given information about individual learners. Their instructional choices moved closer to those of expert educators, but gaps remained. Genuine personalization is still an open problem, and this work is a way to measure progress on it. The study earned the best technical student paper award at AIED 2026.
The team also studies where personalization goes wrong. A 2024 study found that training AI on student data can reduce its performance for students farthest from opportunity, underscoring the need to further evaluate these systems.
left 70% 0 center #FFFFFF 0 02025 Stanford Data Science CORES Workshop
2024 Open Education Conference Keynote
#FFFFFF 80% center #FFFFFFAI by students, for students
Students reimagine the textbook
The people who know best where learning tools fall short are the students using them. In a Rice course last fall, student teams spent a semester with peer tutors and prototyped AI tools around the problems they saw students hit. Summer interns are carrying that work forward.
See AI by students Student AI orange 0 center 0 60% #FFFFFF 2 center dark teal divider -3px 6px 100% center