As Sefaria continues to explore the possibilities for Jewish texts in the digital age, we are beginning to use AI tools to enhance the searchability of our database and provide content that supports and enriches our users’ exploration of the myriad of texts in the library. We currently have two AI features on the site: Enhanced Topic Pages and the Pirkei Avot Learning Guide. You can learn more about Enhanced Topic Pages below.



We believe in the importance of transparency. We always want to ensure an open channel for dialogue with our users and learn what works and what doesn’t. In this spirit, we welcome feedback about AI on Sefaria.



If you’re curious about how we’re building and sustaining the library, you can also look at (or use) our AI models at any time. Sefaria is an open book, as it were, in everything we do — including our data. To access our Github and API, visit the Sefaria Developers’ Portal.



This is new technology; we will be learning and iterating for the foreseeable future. Our goal, as always, is to improve the user experience on Sefaria, specifically around searchability, and to provide avenues through which users can have meaningful independent experiences exploring Jewish texts.

You will now find AI-generated content on some topic pages. These pages are marked with an information (i) symbol on the top right corner.



Some topic pages include an overview paragraph written by our learning team. On AI-generated topic pages, following this introduction, there is a series of AI-curated sources related to the topic punctuated by AI-generated introductory paragraphs. All of the content received some level of human review.



Your feedback is crucial to our ongoing efforts to ensure accuracy and uphold our commitment to quality on Sefaria. Should you encounter anything that seems amiss, we encourage you to let us know by filling out this short AI feedback form.

Sefaria’s Pirkei Avot Learning Guide was created to make learning this text more accessible for all levels of learners. To build this feature, members of Sefaria’s engineering and learning teams used AI to create a new way of learning and engaging with the text.



The Learning Guide summarizes the various commentaries that approach common themes within each Pirkei Avot mishnah. To accomplish this, each mishnah and its commentaries were fed into Google’s AI model, Gemini, which was asked to find recurring questions that showcase different critical approaches and answers. Gemini then shared the questions, as well as summaries of each of the answers.



Every writer needs a good editor, and so, too, does generative AI. We took all of the content that Gemini generated and fed it to Claude, an AI model developed by Anthropic. We asked Claude to review each commentary alongside the questions and summaries generated by Gemini and give them an accuracy score of 1-10. All of the content received some level of human review, but using these scores, our learning team was able to identify content that needed an especially close look. Anything with a score of five or below received a close examination to ensure accuracy and accessibility for learners.

Are you interested in testing out more experiments from Sefaria? Sign up for our Beta Tester list to be the first to hear about new product features, get the chance to test them, and provide valuable feedback.