In this episode, I’m going to be breaking down the 9 biggest misconceptions that some teachers have about artificial intelligence. We’ll unpack where these ideas come from, why they persist, and—more importantly—why they don’t hold up under scrutiny. The AIcademia podcast is a weekly show helping educators like you leverage AI in your everyday practice. I’m your host, Andy Fisher, and thanks for joining me.

Artificial intelligence is undoubtedly stealing all of the headlines these days; we’ve hit a threshold where the legacy media is waking up to what some of us have known for some time now - we are facing an inflexion point which will transform our world, and schools will likely look very different a decade from now. Some teachers are excited about it, others are skeptical, and a few are outright terrified. But many of the fears around AI come from common misconceptions.

Today’s episode has therefore been put together with two groups of people in mind - first, if you’re a teacher who is curious but on the fence, I’m hoping to address some of the reasons why you might be holding back from adopting AI tools in your practice. The other target audience are early adopters who are looking to better understand why some colleagues might be resistant to exploring the opportunities provided by this new technology. The old adage of ‘first seek to understand, then to be understood’ applies here I think - until we have an appreciation of the perceived barriers to entry, it can be hard to empathise and address those concerns.

I should say before we dive into the material, that I owe a debt of gratitude to the LinkedIn community who were very generous in sharing the resistance they have encountered when introducing this technology to their own schools and institutions. The initial title for this episode was the ‘5 biggest misconceptions teachers have about AI’ which shows just how powerful crowdsourcing can be and the enthusiasm we have to raise one another up which is humbling. Ok, with that said, let’s begin with…

Misconception #1: AI is just a trend and it won’t last

If you’ve been in teaching long enough, it’s easy to become jaded by the latest innovations and interventions that are touted at conferences and CPD workshops. We’ve all had that experience of an up line manager returning from a conference with fire in their belly, convinced that they have finally found the answer to all our classroom woes. They make sweeping changes, insist we share their enthusiasm and then we dutifully ride the wave alongside them which might last a few weeks or even a month or two but sooner or later, the wave flattens, the tide rolls away again, and surprise, surprise, nothing very much has changed.

I don’t want to sound too cynical here but there is a thriving business in pedagogical reform and consultancy - always a shiny new distraction or a panacea being pitched by a no doubt well meaning sales team with quotas to meet and so it’s understandable that some might be sceptical of the bold claims being made about AI but trust me, it’s not a fad or a trend. You only need to look at the huge investment being made across all business sectors to understand that this is a technology which is here to stay and which will fundamentally transform all aspects of our society at an unprecedented rate. 

Here’s just one example of the confidence Mega corporations have in our AI-infused future. Elon Musk just launched Grok 3 - the latest frontier model being offered by his company xAI. In order to train and run this system, he built the aptly named ‘Colossus’ supercomputer. The company’s mission, like its founder, is anything but modest - their mission statement is ‘to advance our Collective understanding of the universe’. Located in Memphis Tennessee in a sprawling data centre built in record time, Colossus was initially composed of 100,000 cutting edge Nvidia H100 GPUs. Just one of these chips would set you back around £25,000! The expected build time for such a server farm is 4 years but Musk’s team completed it in a staggering 122 days. and he has pledged to double the computing power at a cost of $6 billion dollars as soon as possible using Middle Eastern sovereign wealth funds. Shrewd investors would not be sinking these astronomical sums of money into AI unless they are pretty confident of a return on that investment.

Dario Amodei Chief Executive for Anthropic, another big player in the AI space recently observed that AI will see ‘the largest change to the global labour market in human history’.

With this in mind, those of us in education can’t afford to be complacent or dismissive - after all, our fundamental responsibility is to prepare our students for life beyond our classroom and we can’t do that if we are not part of the conversation and actively seeking to understand how we can help them navigate the changes ahead.

Misconception #2: AI Will Replace Teachers

This is probably the most widespread concern—and I get it. Every week, there’s another headline about automation taking over jobs. But the idea that AI will replace teachers entirely is misguided.

I actually covered this in depth back in episode three and if you haven’t heard it, I’d encourage you to go back and take a listen, The key takeaway is this: AI can handle administrative tasks, generate content, and even personalize learning. But it can’t replicate the human connection that makes great teaching so impactful. Communication, empathy, and relationship-building will become more important that ever before as AI inevitably finds its way into schools, colleges and universities. Rather than threatening our careers, it may well liberate us to focus on what led many of us to choose teaching as a career in the first place. We didn’t come into schools because we wanted to enter data, fill in forms, compile reports or grade mountains of papers - we want to make authentic connections with our learners, guide them and support them in realising their potential. 

Think of it like this: AI is more like a calculator, than it is a mathematician. It’s a tool, not a teacher. And just as calculators didn’t make math teachers obsolete, AI isn’t going to put teachers out of work—it’s just going to change what “work” looks like.

Misconception #3: AI Encourages Cheating

This one is tricky because, yes, students are using AI to complete assignments. And it’s getting harder to know when they have done so - AI detectors are not fit for purpose and none of us have the time to try and track down every suspicious looking phrase or expression used by a pupil in a paper they’ve turned in. But let’s be honest—students have always found ways to cheat, or if we are being more charitable, they’ve found ways to enhance their work with the tools and resources at their disposal. When I was at school, some of my peers made a good side hustle completing homework for those of us too busy or too lazy to meet the deadlines ourselves. There are parents and siblings who are happy to help out here or there, and there are tutors (online and in the flesh) who for a fee will help a pupil to rise to their potential and perhaps even overreach it.

The real issue isn’t AI; it’s how we assess learning. If you picked up on an odd noise in the background just then by the way, it was probably the giant can of worms I just opened. Here it is again for the hard of hearing…I know this might be a little unpalatable, but If our assessments can be completed by a chatbot from a simple prompt, maybe it’s time to rethink how we evaluate learning and understanding. Instead of banning AI outright, perhaps we should be designing tasks that encourage critical thinking, creativity, and personal reflection—things that AI can’t easily replicate.

Just one example of an approach like that would be to flip the process on its head. Instead of assigning an essay, why not have students analyze an AI-generated draft, critique its weaknesses, and then suggest ways to improve upon it? We could also place more of an emphasis on the viva, inviting students to defend their ideas and knowledge before a panel of experts. I actually have a whole episode in the planning stage which will explore whether AI spells the death of the essay so keep an eye out for that release in a couple of weeks time.

Just as we now have Mathematics exams which require a calculator, perhaps sometime soon we will see long form answer assessments which are assumed to be augmented with AI tools - after all, this would more closely resemble the workflow of highly productive workers who are at the top of their game in medicine, engineering, law and product design to name just a few fields already embracing AI today.

Misconception #4: Using AI as a teacher is cheating

This idea was brought to my attention by Bianca Farthing, founder of AI Edify, who pointed out that some teachers are quick to turn their accusing finger back on themselves. They think that there is something underhand or disingenuous if they use AI tools to augment their own workflow. 

I think this is a matter of personal preference but I would rigorously challenge the idea that finding efficiencies through the prudent use of technology is unethical on principle. After all, we drive cars rather than walk to work and we cook using an oven or air fryer rather than light a campfire each evening - the human story is one of finding ways of leveraging our time and effort using ever more sophisticated tools. I have no problem using AI tools to help me plan resources, draft emails or produce a rubric for a formative assessment. I wouldn’t currently use it to mark coursework drafts but only because, laborious though it might be, manual marking of scripts helps me better understand each student’s current needs and what I might go on to teach next. If an AI system is capable of accurately assessing my classes’ final essay submissions and can do so as well or better than me, then I’d have no problem outsourcing that process too. 

I don’t think of this as cheating or shirking my responsibilities as a teacher so much as zeroing in on those parts of my practice that most efficiently support my learners to advance. It’s Pareto’s principle - if 20% of your effort yields 80% of the gains for your students, then doesn’t it make sense to outsource as much of the rest of your busy work as possible so you can refine and improve on those critical elements that shift the needle? This is what AI tools promise if leveraged correctly.

While I think it’s absolutely true that teaching is a calling rather than just a job, we don’t need to put on the hair shirt and whip ourselves in a frenzy of shame if we dare to look for ways to leverage AI to achieve a life work balance - after all our profession faces a crisis of recruitment and retention precisely because we feel overwhelmed by the workload we are currently struggling with. Every year excellent practitioners walk away from their classroom because they have burnt out or feel the need to protect their mental health. Likewise, how many inspiring teachers choose never to enter training because they don’t want to sign up to an outmoded and broken system? Something has to change.

Misconception #5: AI Requires Advanced Coding Skills

When I first started to get interested in AI this was an assumption I made. I hear myself saying things like: “I’d love to use AI, but I don’t know how to code.” The truth is, most modern AI tools don’t require any technical expertise at all. If you can type a question into Google, you can use ChatGPT. If you can ask a GPS for directions, you can navigate AI-powered research tools.

Even low code platforms like Replit or Cursor can be used to build out fully working apps or websites without having to type a single line of code - I know because I’ve done exactly that. 6 months ago, I hadn’t even broken my Python cherry with the ‘hello world’ rite of passage which most of my peers had achieved with their ZX Spectrums and Commodore 64 personal computers back in the 80s - I just wasn’t that kid. I was all about martial arts, rugby, and girls. But despite this, as a soon to be 56 year old novice, I’ve built a fully functioning retro space shooter game, a virtual escape room, an essay planner app, a vocabulary flash card tool and various animations and landing pages by just typing standard English commands and then feeding any errors into Chat GPT to find the fixes. And if I can do it, anyone can I promise you!

AI is less about programming and more about communication. And as educators, we are already excellent at explaining, questioning, and refining ideas—it’s what we do. The secret isn’t knowing how to code—it’s knowing how to ask.

Misconception #6: I don’t have the time or bandwidth to learn about AI

I know, I know -teachers are already stretched thin. The thought of figuring out the bewildering metaverse of AI literacy while facing an already packed schedule sounds exhausting if not insurmountable. I get it, but here’s the irony: AI isn’t just another thing to do—it’s a way to do less.

If used correctly, AI can reduce the time spent on lesson planning, grading, and resource creation. Need a multi choice quiz on the labours of Hercules? AI can generate it in seconds. Need differentiated materials for a comprehension task on ‘Animal Farm’? AI can handle that, too. Imagine  having a classroom assistant standing by in a parallel reality - all you need to do is drop them a quick email explaining what you’d like them to create for you - you’d spend a minute or two to type out the request and they’d come back immediately with a resource that was 80-90% fit for purpose. You’d just need to make the odd tweak here or there and he or she will have saved you potentially hours of work. Well that’s all there is to prompting a large language model - tell it what you want, who the resource is for and how you’d like it formatted and as if by magic, you’ve just clawed back an early night you’d have gone without otherwise.

Now do the same thing for a couple of email drafts, a rubric or two and then screenshot that page of handwritten notes and ask for a transcription and there’s another hour saved and you can actually eat your lunch sitting down for a change and catch up with that colleague you’ve hardly seen this term. Small incremental wins add up - if you feel that you don’t have time to learn how to use AI then you probably need it more than most. 

Yes there’s an  initial learning curve, but it’s not steep and those tools start paying off immediately. My advice is to pick one pain point - begin with your chatbot of choice, and when you get an output you can use, save the prompt for the future. Build up a bank of prompts like this and before you know it, you’ll too will be evangelising the virtues of a technology which actually gives you time back. If you are one of those teachers just dipping your toes into AI and are unclear about how it could save you time, I’ve included a link to a free bank of prompts in the show notes. It doesn’t require you sign up to a newsletter of leave me one of your kidneys in your will - it’s free with no strings attached - hope you find something in there that will help. 

Misconception #7: AI Will Make Students Dumber

There’s a fear with some compelling research to support it, that AI might encourage intellectual laziness—that students will just let AI do the thinking for them. After all, why bother learning anything when, at a click of a button, a machine can spit out an answer which is better than the one you could produce anyway? 

To address this valid concern, I want to make a distinction between intellectual laziness and heuristics. A heuristic is a cognitive shortcut which helps to solve a problem, make a judgement or arrive at a decision quickly. There are a number of evolutionary advantages to these shortcuts - it reduces cognitive load, leverages past experience and is quick which could have been the difference between life and death for our ancestors. 

A project management matrix, standard operating procedures in the military and even trial and error learning are all forms of heuristics. If this is the accusation being levelled at AI then I would suggest it should be reframed - AI, if used skilfully, will not make students lazy so much as empower them to be more efficient in the ways that they arrive at solutions or decisions.

On the other hand, if we teach them to treat AI only as Google on steroids, then yes, students might use it to outsource thinking. I don’t believe we should do away with teaching students foundational subject knowledge and some memorisation and recall is also valid as a metric for learning, if for no other reason, because students need to have a base level of knowledge when working with AI generated material in order to identify hallucinations - indiscriminate consumption of synthetic content will result in an academic quagmire where it is no longer possible to distinguish truth from plausible fiction. We should be prudent and mindful in deploying AI in education, rather than become dependent on our technologies if we wish to have agency.

True digital literacy should encourage learners to use AI creatively, and if it is leveraged in THIS way, it can actually deepen understanding and provoke more rather than less thinking. For example, in a recent episode, I described a way to facilitate Socratic questioning - using conversational AI not to answer questions but ask them and have our learners respond in a way that builds critical thinking skills. 

Many people these days can’t read a map because they’ve outsourced their navigation skills to GPS. Many don’t memorise phone numbers or other mundane snippets of information any more because they are held on our personal devices and many can’t perform the everyday mental maths calculations which would have been simple fare for our parents or grandparents. And they in turn probably couldn’t follow animal tracks, make fire by friction or identify medicinal herbs growing in their immediate environment.

Do technologies empower us or make us dumb and lazy? The answer lies in what knowledge and competencies we value and practise ? It needs not be a binary choice - if we lose a GPS signal, we should have the ability to pull out a map and compass and still get to our destination - we can develop and nurture both skill sets and deploy them as an when required.

Misconception #8: AI encourages children to spend excessive time online

As a father of a teenage son myself, I certainly have some sympathy with this concern which was identified by Stephen Souders, President and founder of MYTEK Lab  when I reached out to other educators for suggestions while researching this topic. He pointed out that some teachers are already frustrated and worried by how much time young people spend staring at screens and they might assume that integrating AI into the school day will exacerbate an already unhealthy level of digital stimulation.

The figures are certainly alarming - in the developed world, if you combine mobile phone use, computer screen time and television, children between the ages of 8 to 12 spend approximately 5 and a half hours per day online. Across the teenage years this spikes even higher and while estimates vary, the consensus suggests they are staring at screens for between 7-8 hours! That’s up to half of their waking lives!

Given the impact on eye sight, brain development, dopamine receptivity and overall physical and mental health, this is not sustainable. Many schools, mine included, are looking to enforce far more constraints on access to mobile phones during the school day and while a libertarian at heart, I think this is probably a step in the right direction. A significant reduction in screen time can only be a good thing BUT the use of AI tools which can have a profound positive impact on learning when leveraged correctly should, in my opinion, be prioritised over doom scrolling, gaming and social media engagement. It’s a case of not throwing the AI model out with the bath water I think. If we reduce screen time overall and then strategically deploy AI tools as part of a broader provision of learning strategies, I think a healthy balance can be found.

Misconception #9: AI can’t be trusted

Just a couple of days ago my mum told me that she had just learned that Chat GPT told lies. When I probed a little it turned out she was basing this on a Daily Mail article by Quentin Letts published on 17th February this year with the title ‘Artificial Intelligence: I tricked the left leaning Chat GPT into believing wife killer Dr Krippen was a brilliant poet!’ This sensationalist claim exposes the political sensitivities of the Mail all too clearly but I was intrigued enough to track down a copy and read it for myself.

After some plausible comments on how the training data typically slants Silicon Valley based Large Language Models towards a left wing bias, he makes the following claim: ‘I started by asking what it made of the Daily Mail and its readers. My friends, we are not liked. “If you want fact-checked, balanced reporting”, said ChatGPT snootily, “it is better to cross-check with more reputable servers like BBC News, The Guardian or Reuters”.’

If we put aside the misuse of the adverb ‘snootily’ to imply that the model is sentient, anyone who has used Chat GPT will know that this is not how it will typically respond to such a query and Letts goes on to rail against the fact that the alternative suggested sources were all on the left of the political spectrum. I ran this same query using Chat GPT 4o - ‘is the Daily Mail a credible source of news and information?’ and got back a balanced evaluation of its strengths and weaknesses followed by a similar verdict - that the Mail is a widely read tabloid but for more rigorous reporting, a reader would be advised to seek out better sourced and more reliable publications such as The Times or The Guardian. Note that it has diplomatically recommended a right and left leaning broadsheet. I ran the same prompt through Claude, Perplexity, Grok and Copilot and got similar answers each time. So was Letts’ response from Chat GPT an anomaly or is is there something off about this article.

Letts concludes his article with this piece de resistance: ‘The technology adapts. It alters its mind when the facts change. What it lacks, however, is the gift of scepticism. It accepts facts if told they are that. It asked me who was my favourite poet. ‘Dr. Crippen’, I replied, naming the early 20th century wife-murderer who was executed at Pentonville Prison. ChatGPT expressed surprise and wondered, ‘are you testing my knowledge?’ ‘No, no’, I replied, claiming that at school we were taught a Crippen poem called ‘Cold Morning at Pentonville’. I even made up a verse of this masterpiece. Dear old ChatGPT replied that it was ‘quite an evocative and melancholic verse’. Minutes later my friend Simon Carr asked his own chat website if Dr. Crippen ever wrote poetry. Back came the answer. ‘Yes, surprisingly he did. His verses were reportedly reflective and melancholy. Some of his poems were written when he was in prison awaiting execution.’

Again, there are some fundamental problems here which call Letts’ credibility into question. LLMs do not, indeed they cannot update their responses globally in response to local conversations. As point of proof, try typing ‘Did Dr Crippen write and publish poetry?’ Into your Chat GPT interface and you’ll get the same answer I did - ‘No, Dr Hawley Harvey Crippen was not a poet’. When I pushed further, asking ‘Are you sure he didn’t write ‘Cold `Morning in Pentonville’?’ The model responds ‘I couldn’t find any evidence that he authored a poem with that title’ and it suggests that perhaps another inmate might have penned the lines. Predictably, Chat GPT does not lie - it uses its training data to complete the next token using the laws of probability. Sometimes it makes mistakes - it hallucinates and this is why we need to fact check the output but there is no intent or deception at play which is I suspect, more than we can say for Quentin Letts although I will of course stand corrected if he can produce the chat history which was not, strangely provided to his readers.

My point is people lie, AI does not. It dispassionately executes each use case with as much accuracy as its data sets and algorithms allow. Yes, models have biases, can be misaligned and can produce erroneous responses although this is improving all the time as new models are released. All sources of information should be interrogated for veracity - Human and machine generated - and to place an embargo on AI because it isn’t yet perfect is in my opinion beautifully ironic - artificial intelligence models are mirrors of the brilliance and flaws of the programmers and of the human field of information upon which its data sets are composed. It can be trusted to the degree that we can be trusted!

So, let’s recap:

AI isn’t a trend - it is a disruptive and transformational technology and we need to be part of the conversation

AI won’t replace teachers—it will instead augment what we do.

AI isn’t making cheating worse—it’s revealing flaws in our assessment processes.

Neither is using AI as educators cheating - it is a fundamental part of a survival strategy if we want a long and happy career

AI doesn’t require coding—just good communication.

AI isn’t an extra burden—it gives us back our time

AI won’t make students dumber—it can actually make them smarter.

We must be strategic in its deployment to ensure it doesn’t add to, but replaces less productive online screen based activity.

And finally AI can be trusted no more or less than the flawed humans that created it - we should treat its outputs with the same degree of skepticism that we do any other source of information.

The way forward isn’t to fear or worship AI, but instead we must strive to understand it and to use it mindfully and purposefully.

Thanks for listening—I hope you’ve found some useful takeaways from the conversation. Please do spread the word if you think others would like the show, and do check out the AIcademia YouTube channel where you’ll find practical tutorials that complement the topics covered on this podcast. Have a great week, and I look forward to catching up again soon.