Researchers assigned 34 undergraduates to learn histogram equalization, a technique from digital image processing, through either a dialogue with ChatGPT or a human instructor. Both conditions used short dialogic teaching sessions, roughly three to ten minutes.

The two groups performed comparably on a retention test. On the transfer test, the ChatGPT group scored lower, a difference the authors report as statistically significant. With 34 students learning one technical topic, the experiment establishes a directional result in this setting more confidently than it establishes the size of the effect elsewhere.

The distinction matters. A retention question can ask for an idea again, or for steps close to the ones just practiced. A transfer question changes the surface of the problem. The learner has to recognize that the same knowledge is relevant, decide how to use it, and adapt it without being handed the original path.

A useful test after the tutor is gone

If you use an AI tutor to learn a procedure, try a problem that is meaningfully different from the example you just studied. Do it before asking the AI which method applies. If you cannot decide how to begin, that tells you something a smooth tutoring conversation may not.

Transfer is not equally important for every kind of learning. If the goal is factual recall, accurate retention may be the main target. If you expect to use a method under unfamiliar conditions, recognizing when it applies can be central.

This is the same transfer problem that complicates broad claims about “cognitive fitness” training: improvement on the practiced task does not automatically imply improvement on a more distant capability.

The EEG result is interesting, but not the headline

The researchers also recorded EEG while the students learned. The ChatGPT group showed broadly higher theta power, and some regional gamma differences survived the authors' later corrections. Those findings could sound impressive if translated loosely into “AI made the brain work harder.”

The paper itself is more restrained. The pattern is consistent with greater visually driven cognitive effort during a text-based interaction, not with deeper learning simply because more neural activity appeared. Some post-dialogue EEG results also came from a short resting window and deserve caution.

The behavioral result is cleaner: similar retention, different transfer in this small experiment.

“AI tutor” is not one intervention

A different study can easily produce a different result because the system design changes what cognitive work the student has to do.

A 2025 crossover trial in a Harvard introductory physics course tested a purpose-built AI tutor designed around research-based pedagogical practices. Students scored higher after the AI condition than after an in-class active-learning comparison. The published paper described the learning gain as more than double, but a public 2026 re-analysis of the released data argues that this framing depends on using a pooled pretest baseline. Analyses using each student's own pretest still favor the AI condition, but by substantially less. That re-analysis is not itself peer-reviewed, and neither analysis turns one physics course into a general verdict on AI tutoring.

Other experiments have produced still different results. One 2025 study of unrestricted ChatGPT use as a study aid found worse performance on a surprise delayed test 45 days later. Other experiments on AI summaries and reflective prompts around educational videos found no clear retention or transfer advantage in that setup.

A tutor may explain more or less, make the learner generate an answer before helping, provide hints instead of solutions, require retrieval, vary examples, or deliberately test the same principle under new surface conditions. If transfer matters, those are design choices, not properties that come automatically with a chat interface.

Sources

Ulix app

Guten

Guten is a free Android reader for more than 70,000 public-domain books from Project Gutenberg, with offline reading, search, highlights, notes, and optional Premium features.