It’s simple to get swept up in the hype about artificial intelligence tutors. But the evidence up until now suggests caution.
Some studies have actually discovered that chatbot tutors can backfire because students lean on them too heavily, get spoonfed solutions and stop working to soak up the material. Even when AI tutors are designed not to give away answers, they haven’t consistently produced much better results than discovering the old-fashioned method without AI.
Still, scientists who have produced these doubtful studies haven’t quit hope. Some are still experimenting, attempting to construct much better AI tutors.
One promising idea has less to do with how an AI tutor discusses concepts and more with what it asks trainees to practice next.
A team at the University of Pennsylvania, that included some AI doubters, just recently evaluated this technique in a study of close to 800 Taiwanese high school students discovering Python programming. All the trainees utilized the exact same AI tutor, which was designed not to distribute answers.
But there was one essential difference. Half the students were arbitrarily appointed to a repaired sequence of practice issues, advancing from simple to hard. The other half got an individualized sequence with the AI tutor constantly changing the trouble of each issue based on how the student was carrying out and communicating with the chatbot.
The idea is based upon what educators call the “zone of proximal advancement.” When problems are too simple, students get tired. When they’re too hard, trainees get frustrated. The objective is to keep trainees in a sweet area: challenged, but not overwhelmed.
The researchers discovered that students in the customized group did better on a last test than trainees in the fixed issue group. The distinction was identified as the equivalent of 6 to 9 months of additional education, an eye-catching claim for an after-school online course that lasted only 5 months. The AI tutor’s innovator, Angel Chung, a doctoral trainee at the Wharton School, acknowledged that her conversion of statistical systems was “not a perfect estimate.” (A draft paper about the experiment was published online in March 2026, but has actually not yet been released in a peer-reviewed journal.)
Still, this is early proof that little tweaks– in this case, adjusting the trouble of the practice problems to the student– can make a distinction.
Chung stated that ChatGPT’s actions may already feel extremely personal since they are directly responding to a student’s unique questions. But that level of customization isn’t enough. “Trainees generally don’t understand what they do not know,” said Chung. “The trainee does not have the ability to ask the ideal questions to get the very best tutoring.”
To resolve this, Chung’s group integrated a big language model with a separate machine-learning algorithm that examines how students interact with the online course platform– how they address the practice questions, how many times they modify or edit their coding, and the quality of their discussions with the chatbot– and uses that information to decide which problem to dish out next.
How various students communicate with the chatbot tutor
Source: Chung et al, Effective Personalized AI Tutors by means of LLM-Guided Reinforcement Learning, March 2026 Simply put, customization isn’t practically tailoring descriptions. It’s about customizing the finding out path itself.
That idea isn’t brand-new.
Long before generative AI tools like ChatGPT were created, education researchers developed “intelligent tutoring systems” that tried to do something similar: estimate what a student understood and provide the right next problem. These earlier systems couldn’t produce natural discussions, however they might supply tips and immediate feedback. Rigorous research studies found that properly designed variations assisted students discover substantially more.
Their Achilles’ heel was engagement. Many students just didn’t want to use them.
Today’s AI tools could help deal with that problem. Students might feel more interested in a chatbot that converses with them in a nearly human method.
In the University of Pennsylvania research study, students in the personalized group invested more time practicing, about three extra minutes per problem, adding up to about an hour per module in the Python course, compared with half as much time (a half hour or less) for the contrast trainees. The scientists think these trainees did better since they were more participated in their practice work.
Students’ previous knowledge of a subject affected how well the customized sequencing worked. Trainees who were brand-new to Python acquired more than those who currently had Python experience, who did simply as well with the fixed series of practice problems. Students from less elite high schools also appeared to benefit more.
How students’ background affected results
All trainees had access to the very same AI tutor. The treatment distinction compares a customized sequence of issues trouble instead of a fixed sequence, from simple to hard. Source: Chung et al, Effective Personalized AI Tutors through LLM-Guided Support Learning, March 2026
All the Taiwanese students in this study offered for an optional computer shows course that might strengthen their college applications. Lots of were extremely motivated, with highly educated parents, and lots of currently had prior coding experience.
It’s unclear whether the chatbot would work as well with less inspired trainees who are behind at school and a lot of in need of extra assistance.
One possible service: fusing brand-new and old.
Ken Koedinger, a teacher at Carnegie Mellon University and a leader of smart tutoring systems, is experimenting with using new AI designs to alert remote human tutors who can encourage having a hard time students who are drifting off. “We are having more success,” said Koedinger.
People aren’t obsolete– yet.
Contact staffwriter Jill Barshay at 212-678-3595, jillbarshay.35 on Signal, or [email protected].
This story about AI tutors was produced by The Hechinger Report, a not-for-profit, independent news organization that covers education. Sign up for Proof Pointsand other Hechinger newsletters.
Was this story handy? Leave an idea to support your education reporters.
The Hechinger Report is a nonprofit newsroom powered by reader support
Republish our short articles for free, online or in print, under a Creative Commons license.