A Teacher's Perspective on AI Tutoring
"I was skeptical – until I saw the results in my students"
When Jennifer Thornton first noticed several of her Year 7 students mentioning an "AI tutor" during mathematics class, her initial reaction was concern bordering on alarm.
"My immediate thought was: are students just getting answers from AI and not actually learning?" recalls the 34-year-old teacher, who has been teaching mathematics and science at a state secondary college in Melbourne's northern suburbs for nine years.
"We'd already had problems with students using ChatGPT to complete assignments without understanding the content. I thought this was going to be another version of that problem."
But over the course of Term 3, Thornton began noticing something unexpected: the students who mentioned using Emma – an AI tutoring application their parents had purchased – weren't just getting better grades.
They were demonstrating deeper understanding, asking more sophisticated questions, and showing dramatically improved problem-solving approaches.
"I had to reconsider my assumptions," she admits, sitting in her classroom after school, surrounded by student work samples she's been analyzing. "This wasn't students bypassing learning. This was students actually learning more effectively."
The Traditional Homework Problem
Thornton's concerns about AI assistance were rooted in a fundamental challenge that every teacher faces: the homework dilemma. Teachers assign work to reinforce classroom learning, but students complete it at home without access to immediate, expert guidance.
"The reality is that homework often reinforces mistakes," Thornton explains, pulling out examples from her filing system. "A student misunderstands a concept in class – maybe they were distracted for thirty seconds when I explained a crucial step, or they were too embarrassed to ask a question. Then they go home and practice that misunderstanding twenty times on their homework.
By the time I see their work the next day and can provide correction, they've deeply entrenched an incorrect method."
This problem is particularly acute in mathematics, where concepts build sequentially on previous understanding. A student who doesn't grasp equivalent fractions will struggle with fraction operations.
A student who makes consistent errors in algebraic manipulation will find quadratic equations impossible.
"I have 28 students in each class, and I teach five classes," Thornton notes. "That's 140 students across Years 7 to 10. I'm committed to providing feedback, but I'm marking homework for 140 students while also planning lessons, attending meetings, managing administrative tasks, and contacting parents.
The feedback loop is necessarily delayed, sometimes by days."
Australian education research has consistently identified this feedback gap as a significant challenge. Teachers work under substantial pressure – recent surveys show increasing workload stress and concerns about adequacy of support – while students need immediate guidance during the actual learning process.
Observing the Difference
The students using Emma stood out in Thornton's classes within weeks. She noticed the pattern first with Liam, a Year 8 student who had historically struggled with mathematics and regularly submitted homework filled with errors.
"Suddenly, Liam's homework quality jumped dramatically," Thornton recalls. "Not just in accuracy, but in approach. He was showing his working systematically, catching his own errors, and when he did make mistakes, they were genuine conceptual confusion rather than careless procedural errors."
Intrigued, Thornton asked Liam about the change. He explained that he'd been working with Emma each evening, showing the AI his homework and working through problems step-by-step. Crucially, Emma wouldn't simply provide answers – instead, she would identify where Liam's thinking went wrong and guide him to correct his own approach.
"What struck me was that Liam could articulate his mathematical thinking more clearly," Thornton observes. "In class, if I asked him to explain his problem-solving process, he could actually do it. That suggested he wasn't just copying steps – he was genuinely understanding the concepts."
Over the following months, Thornton identified six students in her classes using AI tutoring applications, with four specifically using Emma. All six showed marked improvement, but the changes went beyond simple grade increases.
"These students were making different types of mistakes," Thornton explains, showing examples from her assessment records. "Before, I'd see random errors, incomplete work, and evidence that students had given up when confused.
After they started using AI tutoring, I saw work that was complete, systematic, and where errors occurred, they were sophisticated errors – the kind that indicate a student is engaging with complex concepts and making reasonable but incorrect inferences."
The Classroom Impact
Perhaps most significantly, Thornton noticed that students with AI tutoring support were more confident and engaged during class instruction.
"Sophie used to sit silently in the back corner, never raising her hand, never asking questions," Thornton shares. "She has dyscalculia, and mathematics lessons were clearly anxiety-inducing for her. But in Term 3, she started participating.
She'd raise her hand to ask clarification questions. She'd offer answers – sometimes wrong, but that's part of learning. She'd even help other students occasionally."
When Thornton contacted Sophie's parents, they explained that their daughter had been using Emma for daily homework support. The AI's patient, non-judgmental approach had gradually reduced Sophie's mathematics anxiety, and the nightly success experiences had rebuilt her confidence.
"As a teacher, that's the outcome we dream about," Thornton says emphatically. "We want students to feel confident and capable. We want them to engage with challenging material without fear. If technology can help achieve that, I'm supportive."
The ripple effects extended beyond individual students. Thornton found that her Year 8 class's overall homework completion rate increased from approximately 65% to 87% during Term 3 – a change she attributes partly to several students' AI tutor use creating a positive peer effect.
"When students see their classmates succeeding and gaining confidence, it's motivating," she notes. "Several students who weren't using AI tutoring started putting more effort into homework because they could see their peers achieving results. The classroom culture shifted toward viewing homework as valuable rather than just busywork."
Addressing Concerns
Thornton emphasizes that her support for AI tutoring comes with important caveats. She remains vigilant about ensuring students are actually learning rather than just obtaining answers, and she's adjusted her assessment strategies accordingly.
"I've incorporated more in-class problem-solving where students explain their thinking aloud," she explains. "I use exit tickets where students must complete a problem independently before leaving class. These strategies help me verify that understanding demonstrated in homework is genuine."
She's also observed differences in quality among AI tutoring tools. Some applications, she notes, simply provide worked solutions that students can copy. Emma and a few similar applications appear genuinely pedagogical – guiding students through reasoning processes rather than replacing their thinking.
"The distinction matters enormously," Thornton stresses. "Technology that does students' thinking for them is harmful. Technology that scaffolds their thinking and provides immediate feedback as they work through problems themselves can be tremendously beneficial. It's essentially providing what a skilled tutor sitting beside them would provide."
Cost equity is another concern Thornton takes seriously. As a teacher at a school where approximately 40% of families experience financial hardship, she's acutely aware that educational advantages for wealthier families can deepen existing inequalities.
"When I first learned these students were using paid AI tutoring, I worried about creating a two-tier system where students whose families can afford it get support and others don't," she admits. "But then I looked at the pricing – Emma costs about $30 monthly, which is far less than traditional tutoring at $60 to $100 per hour. It's not free, but it's substantially more accessible than alternatives."
Thornton has begun recommending AI tutoring options to parents who inquire about homework support resources, particularly for families who express concern about tutoring costs. She provides information about several applications, including their different features and pricing, so parents can make informed choices.
"I'm not endorsing specific products," she clarifies. "But when parents tell me they want to get their child tutoring but can't afford $400 per month, I feel responsible for informing them about lower-cost options that might help. That's part of supporting my students holistically."
Complementing, Not Replacing
Thornton is emphatic that AI tutoring supplements but cannot replace classroom teaching and human connection.
"I provide explicit instruction in mathematical concepts, I create a classroom community, I identify and respond to individual learning needs in real-time, I connect mathematical concepts to real-world applications and other subjects," she lists. "AI tutoring doesn't do those things. What it does is provide practice support and immediate feedback when students are working independently – something I simply cannot do for 140 students simultaneously every evening."
She draws an analogy to other educational technologies that teachers initially viewed with suspicion but eventually embraced as valuable tools.
"When calculators became widely available, teachers worried students wouldn't learn basic arithmetic," Thornton notes. "When internet research became standard, teachers worried about plagiarism and misinformation. We learned to adapt our teaching to use these tools appropriately while still developing fundamental skills. AI tutoring is another tool – powerful, requiring thoughtful integration, but ultimately capable of supporting learning if used well."
Looking Forward
Thornton has observed that students using AI tutoring are developing stronger metacognitive skills – the ability to think about and regulate their own thinking and learning.
"These students are getting practice in recognizing when they're confused, identifying specifically what they don't understand, and seeking clarification," she explains. "Those are crucial lifelong learning skills. They're learning to be self-directed learners who can identify and address their own knowledge gaps."
She's also noticed that the availability of AI tutoring has changed some parents' relationships with homework. Several parents have told her they feel less stressed and guilty about not being able to help with mathematics beyond their understanding.
"One mother told me, 'I can be a mum again instead of a frustrated homework supervisor,'" Thornton recalls. "That's significant. When homework creates family conflict and stress, it undermines its educational value. If technology can make homework a calmer, more productive experience, that benefits everyone."
As Australian schools continue grappling with the challenges of diverse learning needs, teacher workload pressures, and educational equity amid cost-of-living pressures, Thornton believes thoughtfully designed AI tutoring represents a potentially valuable support system.
"I'm not naïve about technology," she emphasizes. "I know there are risks – student over-reliance, potential for misuse, questions about data privacy, concerns about further screen time for children. These are legitimate issues requiring ongoing attention and evaluation."
"But I also see what's happening in my classroom," she continues. "I see Sophie raising her hand confidently. I see Liam explaining his mathematical reasoning articulately. I see students who were falling behind starting to catch up. I see families experiencing less homework-related stress. As an educator, I have to acknowledge those outcomes and remain open to tools that genuinely support student learning."
Thornton pulls out her most recent class assessment results, pointing to the overall improvement trend across multiple measures: homework completion rates, test scores, quality of mathematical reasoning, and student self-reported confidence levels.
"Data doesn't lie," she says simply. "Whatever my initial skepticism – and it was substantial – the evidence from my own classroom suggests that quality AI tutoring is having positive effects for students who use it. My job is to support student learning through whatever effective means are available. If that includes AI tutoring as one component of a comprehensive support system, I'm willing to embrace it."
She pauses, then adds: "Though I'd still prefer smaller class sizes, more specialist learning support staff, and better teacher-to-student ratios. Technology is helpful, but adequate funding for education would help even more."
It's a characteristically practical perspective from a teacher who has learned to balance idealism about what education should be with pragmatism about what actually helps students succeed within current constraints. As AI increasingly enters educational spaces, Thornton represents a growing cohort of educators approaching these tools not with blanket acceptance or rejection, but with thoughtful evaluation based on observable student outcomes.
"Ask me again in a year," she concludes with a slight smile. "I'm continuously learning, just like my students. But right now, based on what I'm seeing in my classroom, I'd say quality AI tutoring is a net positive when used appropriately as part of a broader educational support system. I never thought I'd say that, but here we are."