Artificial intelligence is becoming a normal part of how students study. Children can now ask an AI tool to explain a difficult paragraph, simplify unfamiliar vocabulary or help them understand a complicated topic in seconds.
But does easier access to information actually mean better learning?
A recent study involving Microsoft researchers and secondary schools in England suggests the answer is more complicated.
The research, published in Computers & Education, compared how 14- and 15-year-old students learned from texts when using a large language model (LLM), taking traditional notes, or combining both approaches. The study involved 405 students across seven secondary schools in England.
The results offer an important lesson for parents and educators across Europe:
AI can make difficult material easier to understand, but children still need to do some of the learning themselves.
What the Microsoft Study Found
The students were asked to study two history texts. Depending on the group, they either used an LLM chatbot, took notes, or combined the chatbot with note-taking.
Their comprehension and retention were then tested three days later.
The results were revealing.
Students who took notes performed better on comprehension and retention than those who relied on the LLM alone. Combining LLM use with note-taking also produced better outcomes than using the LLM alone.
At the same time, many students actually preferred the AI experience.
They considered the LLM more helpful and easier to use. Students said that the chatbot helped make complex material more accessible and reduced the mental effort required to understand it.
Note-taking, however, encouraged deeper engagement and helped students remember the material.
In other words, the technology was good at helping students get into the material.
Traditional learning activities were better at helping some of that knowledge stay with them.
That distinction may become one of the most important ideas in education in the AI era.
The Problem With Making Learning Too Easy
When a student encounters a difficult text, the natural reaction is to look for help.
Before AI, that might have meant asking a teacher, searching for an explanation online or discussing the material with another student.
Now, an AI assistant can provide an explanation immediately.
That is incredibly useful.
But there is a risk.
If the AI does too much of the cognitive work, students may understand the answer without developing the ability to reach it independently.
The Microsoft research reflects this tension. Students appreciated the efficiency of the LLM, but traditional note-taking produced stronger results for comprehension and retention.
This suggests that the best educational use of AI may not be to remove difficulty.
It may be to make difficulty manageable while keeping the student intellectually involved.
AI Works Best as a Learning Partner
This changes how parents should think about AI tutoring.
The question is not:
“Can AI teach my child?”
A better question is:
“What should AI help my child do?”
For reading and language learning, the answer could include explaining unfamiliar vocabulary, asking questions about a text, providing additional examples, helping a child practise or offering feedback.
But the child should still have to read, speak, write, remember and explain.
Microsoft’s broader 2025 review of empirical research reached a similar conclusion. The researchers found that AI can provide valuable explanations and personalised support, but warned that general-purpose AI can also reduce engagement, encourage overconfidence and interfere with the development of higher-order thinking if it bypasses the effort required for learning.
The report recommends using GenAI as a supplement to traditional learning, with teacher guidance and design features that encourage active engagement.
Why Human Tutors Still Have a Role
This is where human tutoring becomes particularly relevant.
An AI system can explain a word.
A tutor can notice that the child understands the word when reading but cannot use it when speaking.
An AI can generate a grammar exercise.
A tutor can recognise that the child already understands the grammar but lacks confidence.
An AI can analyse an answer.
A tutor can ask why the child chose it.
That human interaction matters because learning is not simply about transferring information. It also involves motivation, confidence, attention and the ability to recognise when something has not really been understood.
Microsoft’s review notes that research continues to show the importance of human connection in learning and that students often prefer human tutors as a trusted source of information.
A More Realistic Model for European Families
For families in Europe, the future is unlikely to be either fully human or fully AI.
A more realistic model is a combination.
A child might use AI for additional practice during the week, use digital tools to explore vocabulary or ask questions about a reading passage, and then work with a teacher or tutor on communication, comprehension and areas where they continue to struggle.
This approach is particularly relevant for children learning English.
European children often grow up in multilingual environments. A child may speak one language at home, another at school and learn English as an additional language.
Their needs can therefore be very different even when they are the same age.
One-to-one tutoring can provide the human flexibility that automated systems cannot always offer.
Platforms such as Khan Academy, Duolingo, Preply, LearnLink https://learnlink.com/tutors are part of this broader shift toward personalised language learning.
The point is not that human tutoring is always better than AI, or that AI is better than traditional teaching.
The evidence increasingly suggests that the combination can be more useful than either approach on its own.
What Parents Should Take From the Research
The Microsoft study offers a relatively simple lesson.
Parents do not need to keep children away from AI.
But they also should not assume that an AI-generated explanation equals learning.
Instead, they can encourage children to use AI in ways that require active participation:
- Ask the AI to explain a difficult concept rather than complete the assignment.
- Ask for questions about a text rather than ready-made answers.
- Use AI to practise vocabulary, then ask the child to use the words independently.
- Encourage children to take their own notes.
- Ask them to explain what they learned without looking at the AI conversation.
These small differences can change AI from an answer machine into a learning tool.
The Real Test of AI in Education
The most important question about educational AI is not how quickly it can provide information.
It is what happens after the screen is turned off.
Can the child remember the idea?
Can they explain it in their own words?
Can they recognise the same concept in a new text?
Can they solve a similar problem without assistance?
The Microsoft research suggests that these outcomes still depend on active learning. AI can make difficult material more accessible and increase student interest, but traditional learning activities remain important for deeper comprehension and retention.
For parents, this means the best educational technology may not be the one that does the most for a child.
It may be the one that helps the child do more for themselves.
That is also where personalised tutoring can complement AI. Technology can provide additional practice and explanations, while a tutor can help a child turn that information into real language skills, confidence and independent learning habits.
As Europe moves further into the AI era, the goal should not be to replace teachers with machines.
It should be to use technology wisely enough that children have more opportunities to learn—and more reasons to think for themselves.
