A look at Newark Public Schools' test of Khan Academy's Khanmigo, one of the first in the US, and worries the chatbot handles too much of the students' thinking
Natasha Singer / New York Times : Twitter: @megankstack and @natashanyt Twitter: Megan K. Stack / @megankstack : Is it my imagination or does Newark often get singled out for experimental school programs? https://twitter.com/... Natasha Singer / @natashanyt : Newark is at the forefront of public schools nationwide that are trying to distinguish the practical use of new A.I.-assisted tutoring bots from their marketing promises. The results so far have been mixed. https://www.nytimes.com/...
Context & Ripple Effects
Newark Public Schools is one of the first US districts to test Khan Academy's Khanmigo tutoring bot in real classrooms, following Khan Lab School's earlier trial of the same chatbot. The district joins a small set of public schools probing AI in class — including a New York City girls' school that had students critique ChatGPT-generated lesson plans — and its early verdict is mixed: the bot risks doing too much of the thinking it is meant to scaffold.
First-order effects
- Newark teachers now have to police where tutoring ends and answer-giving begins, since the report flags Khanmigo handling too much of students' reasoning during lessons.
- Khan Academy gets its first large-district data point on whether simulated one-on-one tutoring holds up outside its own lab school.
Second-order effects
- Districts watching Newark — including ones that had raced to block chatbots before later encouraging their use — gain evidence for whether to adopt or reject AI tutors, with mixed results arguing for cautious pilots over rollouts.
- Rival ed-tech vendors selling AI tutoring will be pressed to show guardrails against over-assistance, since 'the bot thinks for the student' becomes the standard objection.
Third-order effects
- If districts keep running these tests, procurement decisions will hinge on measured learning outcomes rather than vendor marketing promises — a shift from hype-driven adoption toward evidence-gated contracts.
- The pattern points toward AI tutoring becoming a supervised classroom tool with defined limits on assistance, rather than an open chatbot students consult freely.
The trend: US public schools are moving from blocking AI chatbots to structured classroom pilots, with districts like Newark supplying the first real-world evidence on whether AI tutors help or shortcut student thinking.