Two years into the mainstream AI era, education is still figuring it out. Some schools have banned ChatGPT. Others have built entire curricula around it. Universities are rewriting assessment policies on the fly. And students are using AI whether anyone approves or not.

The reality, as usual, is messier than the hot takes suggest. Some applications of AI in education are genuinely impressive. Others are solutions looking for problems. Here’s an honest assessment of where things stand.

What’s Actually Working

Personalised tutoring. This is the strongest use case. AI tutoring systems can adapt to a student’s pace, identify knowledge gaps, and provide explanations in different ways until something clicks. Tools like Khan Academy’s Khanmigo show what’s possible — a patient, always-available tutor that doesn’t get frustrated when a student asks the same question for the fifth time.

The evidence is promising. Several studies have shown measurable improvements in student outcomes when AI tutoring supplements classroom teaching. The key word is “supplements.” AI tutoring works best alongside human teachers, not instead of them.

Administrative efficiency. Behind the scenes, AI is saving teachers time on tasks that have nothing to do with teaching. Generating rubrics, drafting parent communication, summarising student progress reports, scheduling — these are legitimate time sinks that AI handles well.

Teachers in Australia spend a significant portion of their time on administrative work. Anything that shifts even a fraction of that time back to actual teaching is valuable.

Language learning. AI conversation partners are surprisingly effective for language practice. Students can practice speaking and writing in a target language without the anxiety of making mistakes in front of peers. The feedback is immediate, and the system adjusts to the learner’s level.

Apps like Duolingo have integrated AI conversation features, and dedicated language learning tools are emerging that use AI to create immersive practice scenarios. For a country as multicultural as Australia, this has real potential.

What’s Not Working

AI-generated content replacing teaching. Some institutions have experimented with using AI to generate lecture content, reading materials, and even course structures. The results have been mixed at best. AI-generated educational content tends to be generic, lacking the depth and contextual relevance that good teaching provides.

Students can tell the difference. A lecture that feels like it was assembled from generic summaries lacks the anecdotes, the tangents, and the real-world connections that make learning stick. The personality of a great teacher isn’t something you can generate with a prompt.

AI detection tools. The arms race between AI writing and AI detection has been a mess. Tools like Turnitin’s AI detection feature have produced false positives that accused honest students of cheating, while sophisticated AI users can easily bypass detection.

The fundamental problem is that detecting AI-generated text is unreliable with current technology. Research from multiple universities has shown that detection tools struggle with accuracy, particularly for non-native English speakers whose writing style may be flagged incorrectly.

Many Australian universities have moved away from relying on detection tools and are instead redesigning assessments to be AI-resistant — focusing on oral presentations, process portfolios, and in-class work.

Replacing teacher judgment with AI grading. Automated grading for objective questions (multiple choice, fill-in-the-blank) has worked for decades. But extending that to essays, creative work, and complex problem-solving? Not yet.

AI grading systems can evaluate structure and grammar, but they struggle with originality, nuance, and the kind of creative thinking that education should be developing. Over-reliance on AI grading risks optimising for writing that scores well by the algorithm rather than writing that demonstrates genuine understanding.

The Bigger Questions

Beyond what works and what doesn’t, there are deeper issues that education systems are grappling with.

Equity. Students with access to better AI tools and the digital literacy to use them effectively have an advantage. This risks widening existing gaps between well-resourced and under-resourced schools. If AI becomes essential to learning, ensuring equitable access isn’t optional — it’s a moral imperative.

What we’re actually assessing. AI forces a confrontation with what education is for. If a student can produce a competent essay with AI assistance, what are we actually testing? Knowledge? Writing skill? Critical thinking? The ability to use tools effectively? These are legitimate questions without easy answers.

Teacher training. Most teachers haven’t been trained to integrate AI into their practice. They’re figuring it out on their own, which leads to inconsistent and sometimes counterproductive approaches. Australia’s education departments need to invest seriously in professional development for AI in education. Not one-off workshops, but sustained, practical training.

What Should Come Next

The institutions getting this right share a few characteristics.

They’re treating AI as a tool to be taught, not a threat to be banned. Students will use AI in their careers. Teaching them to use it responsibly and critically is arguably more important than trying to prevent its use.

They’re redesigning assessments rather than just policing them. The essay-as-default assessment was already questionable before AI. Now it needs a genuine rethink.

They’re investing in teachers, not just technology. The best AI tool in the world is useless if the teacher doesn’t know how to integrate it effectively.

And they’re being honest about what they don’t know. AI in education is still early. The evidence base is growing but incomplete. Institutions that stay flexible and keep experimenting — while being transparent with students about the process — are better positioned than those that commit rigidly to any single approach.

The Bottom Line

AI in education isn’t a simple good-or-bad story. It’s a tool with genuine strengths and real limitations, deployed in a system that was already stretched thin. The challenge isn’t whether to use AI in education. It’s how to use it in ways that genuinely serve learning — and that requires more thought, more investment, and more honesty than most of the current debate allows.