The third equation of motion had Zohan Ali confused. He knew the formula, but the derivation was another story. Somewhere between the speed-time graph, the area of a trapezium and the algebra that followed, he was losing track of how the equation came together.

Ali, a Class 9 science student, was studying kinematics. He could have memorised the derivation from his physics textbook, but he wanted to understand it. So, he turned to Artificial intelligence (AI).

His question was fairly specific. He told the chatbot he was a ninth-grade student following Pakistan's board syllabus and asked it to explain the derivation as a private tutor would.

It started with the graph. The area beneath it represented the distance travelled. From there, it explained the trapezium formula, substituted the variables and worked through the remaining algebra.

“Seeing each line of algebra laid out clearly helped me grasp the logic instead of just memorising the proof,” he says.

This wasn't Ali's first attempt at studying with AI. He had heard classmates discussing it back in class 8, when he was struggling with physics numericals. Private tuition wasn't always answering his questions, and he wanted to see whether a chatbot could explain things differently.

He now uses the chatbot regularly when he gets stuck. Sometimes he asks it to explain a concept. Sometimes he uses it to prepare for tests. And sometimes, with homework piling up, he admits that getting an answer without working through the question himself can be tempting.

“Honestly, yes,” he says when asked whether he ever uses AI for quick answers. “When I have multiple homework assignments due the next morning alongside heavy board exam prep, it's tempting to use AI to find quick answers.”

There is a difference between asking a chatbot to explain a physics equation and asking it to finish the homework. For Ali, both possibilities are only a prompt away.

A tutor that never closes

Fardeen Saleem has two sets of deadlines to worry about. There are campaigns and other responsibilities at the private company where she works as a brand manager, and then there is the coursework for her MBA in marketing. Keeping up with both leaves her with limited time to sit down and work through difficult topics.

She remembers spending considerable time going through different resources whenever something in her coursework wasn't clear. These days, she often starts by asking ChatGPT. If the first explanation doesn't help, she asks it to simplify the answer or explain the same concept through an example.

“Earlier, I would spend considerable time searching through different resources just to understand a concept,” she says. “It has made my learning process more interactive.”

Much of her coursework involves applying marketing theories to business situations. She uses AI to understand concepts such as consumer behaviour, market segmentation and SWOT analysis, and to see how these frameworks might work in practice. It also helps her organise material before she begins an assignment.

Fardeen uses several platforms, depending on what she needs. ChatGPT is her regular choice, while Claude helps with longer documents and Perplexity with finding information that includes references. She also searches Google Scholar for academic papers. She is aware that AI can get things wrong, particularly when it comes to references, and says she checks information against original sources.

The ability to ask questions repeatedly is something Ali has also found useful, although his concerns are rather different from Fardeen's. Before a Biology test, the class 9 student was having trouble with topics including cellular respiration and photosynthesis. He asked AI for simpler explanations and then for a ten-question practice test, including explanations of the answers he got wrong.

Ali says he scored above 85 per cent in the unit test, having previously struggled to score between 65 and 70 per cent. He credits the preparation with helping him understand the material, though there were other factors that may have contributed to his result.

For Fardeen, who has to fit university work around her job, the biggest advantage is not having to spend as long searching for an explanation that makes sense. She can ask another question, compare two concepts or request a different example before returning to her coursework.

AI enters the classroom

When Tahir began teaching problem-solving to his Class 8 students at Daanish School, two concepts were proving particularly difficult: pseudocode and algorithm optimisation. Explaining them through the textbook wasn't getting him very far, so he decided to show the students how different AI models approached the same problem.

“I tried different AI models for the same problem in front of them and the results were clear,” he says. He asked the models to optimise a path between two points, allowing the students to see how the problem could be approached. “Students were amazed to see the results.”

Tahir teaches computer science and robotics to classes 6 and 8. A software engineering graduate, he is also pursuing a master's degree in Artificial Intelligence. He recently joined teaching and has been experimenting with ways to make difficult concepts easier for his pupils to understand.

Being a boarding school, Daanish does not allow students to keep mobile phones with them all the time. But classrooms have smart screens, and tablets are available for accessing Google Classroom and searching for information.

“I have installed ChatGPT in their screens on their request,” Tahir says. Students either look up information themselves or ask their house masters for help.

Language is another difficulty he encounters. Some pupils are comfortable reading English textbooks, while others struggle with the wording. Tahir says he has noticed that students understand topics more quickly when they are given examples from everyday life.

“I have noticed that they pick up very quickly if they are given real life examples,” he says. Students who struggle with English also use AI to find Urdu meanings and translations of textbook passages.

Alongside general-purpose chatbots, teachers are also exploring AI applications designed specifically for education. One such platform is Gauth, which allows students to photograph or type in academic questions and receive step-by-step explanations. Beyond solving problems, its Quizify and Flashcard features help turn study materials into revision exercises, while TLDR produces shorter summaries of lengthy content.

The platform also includes an AI Tutor for interactive explanations and Gauth Atlas, which uses illustrations to explain topics such as biological structures and scientific processes. These features give students different ways to revisit difficult lessons, although understanding the explanation remains more important than simply obtaining the answer.

The use of AI in education is also gaining institutional attention. Under an agreement with the government, Gauth is set to provide more than 4,000 free subscriptions to teachers across Pakistan, including those associated with the government-backed Daanish Schools. Tahir says teachers at Daanish School Islamabad have begun exploring the platform and how it might be used in their classrooms.

AI is also being used beyond regular lessons. Students have begun preparing speeches for school functions and assemblies with its help, working alongside their teachers. Tahir says their curiosity sometimes leads them to look up unusual questions and then put those same questions to their teachers.

For him, AI has become another resource in the classroom, but he is wary of teachers depending on it for everything. Preparing quizzes and assignments entirely through AI, he argues, can overlook the different abilities of students sitting in the same class.

“Teachers can understand that not every class has all bright students,” he says. “AI should be there just for supporting and not for doing everything.”

When the answer comes too easily

Maroof Ahmed was going through his class 10 students' physics essays when he came across a sentence that made him stop reading.

“Electromagnetic induction serves as the quintessential linchpin of modern power distribution, seamlessly facilitating the transition from mechanical kinetics to electrical potential.”

The assignment was about electromagnetic induction, how it powers household devices and one limitation of the technology. Ahmed had expected the students to explain the topic in their own words. The boy who submitted this essay usually wrote in fairly simple English. Suddenly, his work was full of elaborate vocabulary.

Ahmed, who has been teaching physics and chemistry for more than eight years, decided to ask him about it after class. What did he mean by “quintessential linchpin”? Could he explain “flux linkage”?

The boy couldn't. He looked down, embarrassed, before admitting that he had copied the answer from chatbot into a Google Doc. He hadn't even read the whole thing. There were three other assignments due that night.

Rather than give him a zero, Ahmed took him to the staff room and asked him to spend 20 minutes reading what he had submitted. He had to look up the unfamiliar words and then write a summary himself.

Ahmed says he has become accustomed to seeing students turn to AI whenever they encounter a difficult question. He remembers a time when pupils would bring him notebooks filled with attempts at solving a numerical, even if most of those attempts were wrong. These days, he says, some students reach for a chatbot before trying to solve the question themselves.

“The willingness to sit with a problem, try three wrong approaches, and figure it out independently has diminished because they are accustomed to instant step-by-step solutions,” he says.

He finds the problem particularly noticeable when he changes a familiar physics question slightly. Students who have learnt a particular solution can struggle to work out what to do when the numbers or conditions are different.

Kashaf, at university, has faced a similar problem from the other side. She is in the final semester of her software engineering degree and regularly uses chatbots while working with Python and machine learning.

On one assignment, she received AI-generated code that ran without errors. That should have been good news, except she couldn't understand everything the programme was doing.

“AI-generated code can work without me immediately understanding every part of it,” she says.

Kashaf went back to chatbot, this time asking it to explain the code line by line. She ran the programme herself, checked its output and made small changes to see how they affected the results. She has also used AI to understand Decision Tree and Random Forest models, particularly how they are trained and used to make predictions.

“This shows that getting a working solution is not the same as actually learning how to develop one,” she says.

For Ahmed, the challenge is deciding how much a student's submitted work tells him about what the student knows.l

“Traditional take-home written assignments and generic essay topics are no longer reliable indicators of student competence,” he says.

The class 10 student had submitted an essay containing terms he couldn't explain. Ahmed only found that out because he asked him to talk through what he had written.

The students on the other side of the screen

For students at Daanish School, using AI during a lesson is relatively straightforward. There are smart screens in classrooms and tablets available for academic work. But Tahir knows that not every student in Pakistan has access to the same facilities.

He worries about those who have no personal device or reliable internet connection, particularly as AI becomes a more common part of studying. A student with access to a chatbot can ask for another explanation after school, practise questions or revisit a topic as often as needed. Someone without that access must rely on whatever help is available elsewhere.

The difference is not necessarily in the students' abilities, Tahir believes, but in the resources available to them.

The issue becomes more complicated when language is involved. At his school, Tahir has seen pupils turn to AI to translate difficult English passages into Urdu. Some understand the subject but struggle to follow the language used in their textbooks.

For a child who cannot afford regular private tuition, having access to explanations outside school hours could be useful. But that possibility comes with an obvious limitation: the student first needs a suitable device, an internet connection and enough familiarity with the technology to ask useful questions.

Ahmed, who teaches physics and chemistry in Karachi, has also seen students use AI to translate material into Urdu and other languages they understand more comfortably. He considers this one of its practical uses, especially when difficult terminology gets in the way of understanding a lesson.

Yet he is reluctant to treat access to AI as an answer to every learning difficulty. Students still need someone to recognise when they have misunderstood a concept, he argues, and teachers must be able to judge whether the explanation a student has received is correct.

There is also the question of what happens when students begin depending on the technology for work they could have attempted themselves. Ahmed has seen pupils turn to AI almost immediately when faced with unfamiliar questions, rather than first trying to work through them.

The same tool that helps one student understand a difficult English passage may allow another to complete an assignment without understanding it. For teachers such as Tahir and Ahmed, making AI available is only part of the challenge. How students are taught to use it matters just as much.

The real test is what remains

For Ahmed, the way forward involves changing what teachers ask students to do. He wants more oral questioning, classroom exercises and assignments that require pupils to explain how they reached an answer.

While Tahir is equally cautious about allowing AI to take over the teacher's role, Ali has his own way of finding out whether chatbot has actually helped him learn something. After using it to understand a difficult topic, he closes the screen and tries to explain the answer aloud, in Urdu or English, without looking at his notes.

He calls it his “re-explain test”. If he gets stuck, he knows he needs to study the topic again.

Fardeen, who uses AI while studying for her MBA, puts it simply: “There is a difference between using AI to learn and using it to avoid learning.”

For Ali, that difference will matter when he sits his board examinations. He will have to work through the physics questions himself, without a chatbot to take him through the steps.