Getting started with AI

AI moves at speed and keeping up is tough! Mark Anderson returns with a brand-new series on the teacher-specific aspects to consider as you use it.

Back in March 2023, I wrote a piece for RISE Education Magazine on getting started with ChatGPT. At that time, it was a bit of a mixed landscape. Some schools were banning it. Some were embracing it. Many were somewhere in between and not knowing quite what to do.

Three and a half years on, things have moved on considerably, and so here, I explore afresh, with a focus on the five areas I’m asked most about with the schools and trusts I work with.

Which AI tools should I use?

The problem here isn’t so much that the tools teachers might use are inherently bad; it’s about governance, safeguarding and the protections required for using them in a school. Just as we perform risk assessments when we organise a school trip, so we should do the same with the tools and products that handle children’s data.

Using AI tools in school that haven’t been through those checks is known as ‘shadow AI’ – and it’s a big problem facing school leaders today.

The DfE’s guidance on generative AI in education is clear that we should keep personal data out of AI tools and warns that many free tools use what you type in as their training material. It’s equally clear that pupils’ work, lesson plans and even prompts can be copyright material, and pupils’ original work must not be used to train AI without permission or a copyright exception. Keeping children safe in education 2026 also shares that same guidance, so this is a safeguarding issue as well as a data protection one.

For most teachers and colleagues, that means using the tools your school provides, signed in with a school account. The issue is that a lot simply don’t.

How do I write a good AI prompt?

In my original piece, I mentioned GIGO: garbage in, garbage out. These days, I talk about PIPO: pedagogy in, pedagogy out. Put weak pedagogy into a prompt, and AI will mirror it right back at you. If you ask for a lesson plan that caters for learning styles, you’ll get one, even though that thinking was debunked long ago. And if you ask for differentiated resources when what you need is adaptive teaching, you’ll get worksheets sorted by ability.

On prompting and productivity in your use of AI, it’s also helpful to know when to stop. I tell staff to aim for AI getting them around 80% of the way there (as per Pareto’s principle) and then finish the rest themselves. Barry Schwartz’s research on maximising versus satisficing found that people who hold out for the best option end up less satisfied, and more regretful, than those who settle for ‘good enough’.

Then there’s AI slop, Merriam-Webster’s word of the year for 2025: low-quality content produced in bulk by AI. Chantal Shaib and colleagues at Northeastern University found that experts judge text as ‘slop’ when it’s thin on substance, off the point and formulaic. Add to this the tics that give AI writing away, like dashes scattered through every paragraph, “it’s not X, it’s Y” constructions and words like “delve”. Each model has its own favourites too – Claude, for one, leans heavily on “quietly”, “what matters” and “throughline”.

Can I trust what AI tells me?

Not without checking. Generative AI is not a database. It predicts likely next words rather than looking up facts, so whilst it is really useful, it can produce confident content that is completely wrong. This is known as ‘hallucination’.

Bias is harder to spot. AI gains its knowledge by being trained on online content such as from Reddit and Wikipedia. From that, it absorbs assumptions which then appear in the responses it generates. The responsibility for anything we produce rightly stays with us, as the DfE makes clear, so this makes our domain knowledge around pedagogy and content even more important.

Arran Hamilton, John Hattie and Dylan Wiliam’s work on the future of AI in education helps explain why domain knowledge is so important. The experts who use AI have strong pedagogical and content domain knowledge and can therefore understand and extend what they produce. In contrast, the novices who use AI tend to stay novices, because the AI does the work that would have helped to build that expertise.

How can AI help me adapt resources for my learners?

Workload is the daily issue that impacts teachers and their productivity. One area that takes considerable time where AI can help is in the adaptation of resources. Rewriting a source at three levels with a school-approved tool and a good prompt can bring you first drafts in minutes. You then (hopefully, obviously!) simply need to check the accuracy of the content and that the challenge hasn’t been stripped out – thus saving you time with the help of AI.

Something I always keep at the heart of everything I do is to remember: pedagogy first, technology second. It’s important to keep the main thing the main thing.

Is AI making us think less?

Cognitive offloading isn’t new. It’s human nature to want to get from A to B as quickly as possible. We see it all the time with alternative short cut paths worn into the grass across parks. They’re called ‘desire paths’ but the thing is, learning isn’t easy. It requires struggle. So, when we offload our learning to something else, we don’t, well… learn! As Willingham reminds us, memory, after all, is the residue of thought.

For example, a field experiment published in PNAS gave nearly a thousand secondary maths students in Turkey access to GPT-4 for practice. Their practice scores rose, but in the exam without AI, the students who’d used standard AI scored 17% lower than those who had never had it. A version built with teacher-designed hints which wouldn’t give answers away, largely removed that harm.

The DfE added cognitive development to its generative AI product safety standards in January 2026.

Where should I start?

Getting started with AI doesn’t mean using it for everything. TPACK reminds us that good teaching with technology happens where technological, pedagogical and content knowledge meet – and that includes knowing when not to use technology.

So, start small. Choose one thing that really eats into your time. Use a tool your school has approved, check everything it gives you, and try to notice where you’ve saved time.

And remember, sometimes, the best choice will still be to step away from the AI and do it yourself.