As a member of a UK academic committee responsible for reviewing education innovation and safeguarding academic standards, I have spent the past two years watching the rapid evolution of artificial intelligence with both curiosity and caution. But no development has travelled more quickly from technological curiosity to classroom reality than the use of AI—especially ChatGPT—to generate reading-comprehension questions.
Across the UK, from Year 4 classrooms in Newcastle to GCSE revision groups in London, teachers are increasingly experimenting with AI systems that can, in seconds, produce comprehension activities that once required hours of careful planning. The transformation has been swift, sometimes chaotic, occasionally worrying—and absolutely impossible to ignore.
In this commentary, I want to unpack how this shift happened, what it means for British education, and why the debate now unfolding will shape the literacy of an entire generation. Some will argue that “question generation” is a trivial subfield of AI. In reality, it cuts to the heart of how we teach children to read, think, and understand the world.
And that makes it one of the most important discussions of our time.
When ChatGPT first launched in late 2022, most educators were concerned that pupils would use it to write essays. Few predicted that teachers would quickly discover a far more practical application: generating the comprehension questions needed to support daily reading practice.
Reading comprehension tasks are the backbone of English literacy instruction. Teachers routinely create:
factual recall questions
inference questions
vocabulary prompts
multiple-choice quizzes
open-ended analytical tasks
short written responses
scaffolded sentence starters
For decades, these were written manually. A teacher might spend 30 minutes drafting high-quality questions for a single text—and do this daily.
Suddenly, ChatGPT could produce the same set in fifteen seconds.
The speed was irresistible. It freed teachers from repetitive work. It allowed more tailored content. And it arrived at a moment when schools were already facing overwhelming pressure—from OFSTED requirements, staffing shortages, and the need to support pupils whose literacy suffered during pandemic disruptions.
It was, in other words, the perfect storm for rapid adoption.
When asked why they use ChatGPT for reading comprehension, teachers offer three consistent reasons.
A Year 6 teacher in Leeds told our committee that AI tools saved her “at least six hours a week”. For a profession in which evenings and weekends are already swallowed by marking, planning, and data recording, saving six hours is nothing short of transformative.
Producing three versions of a comprehension task—for lower-attaining, middle-attaining, and high-attaining pupils—can be exhausting. AI enables customised sets in minutes.
Teachers can request:
simplified texts
dyslexia-friendly versions
expanded glossaries
culturally responsive adaptations
translations for EAL learners
audio narration support (when combined with text-to-speech)
This creates a level of inclusivity that many schools have long struggled to achieve consistently.
For many educators, these benefits feel like a long-awaited relief.
The moment an AI system becomes responsible for shaping what pupils read, it becomes a de-facto co-teacher. And like any teacher, it can get things wrong.
Sometimes spectacularly.
AI models occasionally invent details not present in the original text, leading to inaccurate comprehension questions. If a story mentions a girl walking her dog, the AI may generate a question referring to a cat. For young children still learning to interpret texts, such errors are harmful.
AI often struggles to match the cognitive difficulty required for SATs, GCSE English Language, or other assessments. Questions may be either too literal or too advanced.
ChatGPT occasionally produces questions that rely on American idioms, references, or assumptions that do not align with British curriculum expectations.
AI reflects the biases of its training data. That means depictions of characters, settings, or moral choices may inadvertently reinforce stereotypes.
Though ChatGPT generates syntactically correct questions, it does not inherently “understand” the learning progression required to build reading competency. It may skip crucial intermediate steps or overemphasise certain skills while ignoring others.
These are not minor issues. They strike at the core of literacy development.
Within our academic committee, we have debated for months a deceptively simple question:
Can an AI model produce reading-comprehension tasks that genuinely support learning, rather than simply “look like” they support learning?
Experts are divided.
They argue that:
AI improves efficiency
teachers retain oversight
models can be fine-tuned to UK curricula
the benefits outweigh the risks
Some even predict that AI-based question generation could lead to more personalised literacy instruction than previously imaginable.
They counter that:
comprehension questions are scaffolds for thought
poor scaffolds produce superficial reading habits
AI lacks a theory of mind or narrative understanding
literacy outcomes may suffer silently over time
To the sceptics, the fact that AI can produce plausible questions is not evidence that it produces educationally effective ones.
I find myself somewhere in the middle.
If AI merely replaces human-written questions with machine-written ones, nothing significant changes.
But the deeper danger is that schools begin to conflate the appearance of comprehension with the experience of comprehension.
A well-crafted question does more than check understanding; it guides the learner toward cognitive processes:
inference
prediction
character motivation
authorial intention
structural analysis
comparison across texts
moral reasoning
AI, by contrast, often produces questions that mimic surface structure without embedding cognitive intention. They may appear correct but fail to cultivate the mental habits that define true literacy.
As one literacy researcher told our committee:
“An AI can generate the shell of a question, but it cannot embed the pedagogy inside it. And pedagogy is where the learning lives.”
This is the heart of the debate.
Perhaps the most surprising aspect of this transformation is how little the general public knows about it. Parents across the UK regularly help their children rehearse reading-comprehension activities at home, unaware that many of those questions were generated by an AI system.
If you ask parents whether they want AI-generated homework, many hesitate. They worry about accuracy, fairness, and screen-mediated education.
But they are rarely asked.
This disconnect between classroom practice and parental awareness raises difficult ethical questions. Should schools be required to disclose when AI plays a role in instructional design? Should parents have the right to opt out? Should AI-generated tasks be labelled?
These questions remain unresolved.
Literacy education is a form of cultural transmission. Through reading, children internalise:
values
social norms
empathy
moral frameworks
identity narratives
When AI systems generate reading passages and questions, they indirectly shape these cultural elements.
This raises profound moral considerations:
Whose voices are represented in AI-generated content?
Whose stories are marginalised or misrepresented?
What cultural assumptions are encoded into the prompts?
Should corporate AI systems have influence over children’s moral development?
For a nation that values educational autonomy, these questions matter.
Many teachers tell us they appreciate AI-generated materials but fear becoming dependent on them. Several expressed concern that new teachers may lose the ability to design effective tasks from scratch.
This echoes the calculator debate in maths education: efficiency comes at the risk of eroding foundational skills.
Some headteachers have floated the idea of “AI-proof teacher training”, ensuring that pedagogical expertise remains intact even as tools evolve. This may eventually become essential.
For all the concerns, it would be a mistake to overlook the enormous potential. Used well, ChatGPT can:
expand access to tailored reading support
help struggling readers build confidence
reduce teacher workload
enable rapid remediation
generate culturally aware materials for diverse communities
allow pupils to practise independently at home
support adults in literacy programmes
strengthen consistency across multi-academy trusts
For pupils with dyslexia, ADHD, or language barriers, AI-generated scaffolds could be life-changing.
But only if we use them thoughtfully.
After extensive research, expert interviews, classroom visits, and review of teacher experiences, our committee has begun sketching out a framework for responsible adoption. It rests on six principles.
AI can support, but never replace, professional judgement.
Schools should disclose when AI-generated material is used, and why.
Questions must be checked for:
factual accuracy
alignment with UK curricula
cognitive appropriateness
cultural sensitivity
bias reduction
Teachers should evaluate not just whether questions are “correct”, but whether they foster meaningful learning.
Teachers need structured training on:
prompt design
bias detection
output evaluation
pedagogical integration
AI must not be used to subtly shape belief systems, cultural interpretations, or political attitudes in reading materials.
These principles are a starting point, not a conclusion.
In the next five years, AI systems may be able to:
generate comprehension tasks mapped precisely to KS1–KS4 progression
analyse pupil responses and diagnose reading difficulties
adapt questions in real time
create personalised reading journeys
annotate texts with cognitive scaffolds
identify gaps in reasoning
offer teacher-assistive explanations
co-design curriculum sequences
Some of this is inspiring. Some is unnerving. All of it demands careful governance.
At present, conversations about AI in schools happen in scattered pockets: a staff room here, a multi-academy trust conference there, a worried parent forum online. What we lack is a national conversation about how AI will redefine literacy.
The stakes are enormous.
Reading comprehension is not merely an academic skill. It influences:
employment
civic participation
political literacy
emotional development
cultural integration
social mobility
If AI reshapes how children learn to interpret texts, it may reshape the future of British society itself.
That debate deserves attention—from policymakers, journalists, parents, and the public.
ChatGPT has become an invisible collaborator in British classrooms. Sometimes it enhances learning. Sometimes it introduces risk. Most of the time, its influence is unnoticed.
But education is too important to leave to chance.
If we want AI to elevate literacy rather than erode it, we must treat reading-comprehension question generation not as a technical novelty, but as a profound educational intervention that requires oversight, transparency, and scholarly vigilance.
The future of reading in Britain will be shaped not only by algorithms, but by the values we embed into their use.
And we owe it to every child to get it right.
