Making the step up from GCSE: using AI to develop precise A-level terminology
How I use examiner reports and AI to identify where familiar GCSE answers need greater precision, depth and application at A level.
Across the CCEA GCE AS1 Principles of Nutrition Chief Examiner Reports (2017–2026), a recurring pattern emerges: candidates can limit the quality of their responses by relying on GCSE-level wording, basic definitions or generalised recall, rather than developing answers through precise AS-level terminology, physiological mechanisms and application.
As a teacher, this is something I want my students to recognise early as they begin their AS studies. They arrive with knowledge they can build on, but they also need to understand how their answers must develop.
For example, “growth and repair” is a familiar description of the functions of protein. At AS-level, I teach them to develop this into the synthesis of new tissue during the growth and development of children and teenagers, and the repair and maintenance of body tissue, for example after illness or surgery.
In my AS1 Nutrition & Food Science classes, we explore these examples during the protein and fat topics, revisit the terminology before the end-of-topic test and then examine their answers afterwards. Looking at where wording or insufficient explanation has cost marks helps make the expectations concrete.
In this post I’ll explore how Gemini Notebook can help collate examiner comments about this transition from GCSE to A level. The purpose is to turn those findings into examples we can discuss, practise and revisit with students, so they become more aware of the precision, depth and application their answers need.
There is an important teaching distinction here. Sometimes an answer is inaccurate. Sometimes it is correct but too general for the question. Sometimes it is underdeveloped, because the relationship between ideas has not been explained. Sometimes it is not applied to the context.
Those difficulties need different responses. Simply telling everyone to “use more scientific words” will only get us so far.
For my example, I will focus on the CCEA AS1 Principles of Nutrition unit in Nutrition & Food Science, using the examiner reports as the source material. I asked Gemini Notebook to analyse the reports and collate comments about GCSE-level recall, specialist vocabulary, depth of explanation and application.
This gave me a focused way to revisit the reports: looking specifically for where students’ existing knowledge needed to develop as they moved to AS level.
The advantage for me was not that AI replaced reading the examiner reports. It was that it could search across several years of reports at once and bring together comments relating to one very specific issue. Instead of manually trawling each report looking for references to terminology, depth and application, I had a starting point that I could then check against the original sources.
I then used the Data Table feature to present the findings side by side in a Word document.
For the fat topic, the output brought together examples involving fatty acid structure, cholesterol, cardiovascular disease and the functions of fat. A table made the differences much easier to analyse than a long written summary.
One example concerned the distinction between describing how trans fats are produced and describing their molecular structure. A student might recall hydrogenation and explain the manufacturing process. However, if the question asks about structure, that explanation does not address the same task. Students need to recognise what the question is asking before deciding which knowledge to use.
Another example concerned the language used to discuss LDL and HDL. Phrases such as “good fat” and “bad fat” can conceal confusion between types of dietary fat and lipoproteins. The teaching opportunity is to ask students to explain the relationship accurately, using the detail expected by the particular question and its mark scheme.
The table also highlighted application. When a question asks about the functions of fat in infancy, students need to connect their knowledge to infants’ requirements. A general list of functions may leave that connection unexplained.
For me, the value of the table is that it gives us something specific to discuss. We can ask: Is this answer inaccurate, insufficiently precise, underdeveloped, or missing the context? That is a more useful conversation than treating every shorter answer as “GCSE wording”.
I then extended the resource by asking Gemini Notebook to generate a question-specific assessment grid for a 12-mark essay on dietary fat, followed by exemplar responses aimed at Bands 2, 3 and 4.
The generated practice question was:
Evaluate the health implications of dietary fats, contrasting saturated and trans fatty acids with unsaturated fatty acids in relation to blood cholesterol levels and cardiovascular disease. [12 marks]
The purpose was to connect the terminology table with a complete answer. Students can learn individual terms, but they also need to see how those terms contribute to an explanation, a comparison and an evaluation.
To construct a classroom grid, I would supply the relevant official CCEA assessment criteria, together with the question and appropriate indicative content, then ask AI to make the descriptors concrete for that question. The official band boundaries and requirements must remain the reference point. Any question-specific descriptions added by AI are teaching adaptations.
This matters because a response does not automatically reach a higher band by including a particular number of technical terms. Accuracy, relevant knowledge, explanation, application and the demands of the command word all need attention. The grid should help students understand those expectations together.
The next step was to request three contrasting essay samples, aimed at Bands 2, 3 and 4, with accompanying AI-generated marking commentary.
The focus was on showing how terminology and depth could develop across the responses: from general statements, through more developed explanations, to a fuller, more precise answer that addresses the question throughout.
These are intended-band exemplars. Their suggested bands and marks need teacher checking, and the accompanying commentary is AI-generated feedback. CCEA’s published examiner reports remain a separate source.
One possible classroom activity would be to give students the three responses without revealing their intended bands. Working with the checked grid and fat terminology table, they could:
Highlight accurate scientific terminology and identify any misleading wording.
Underline explanations that develop a point rather than simply state it.
Identify where the response compares fat types and evaluates their health implications.
Award an appropriate band, justifying the decision with evidence from the response and assessment grid.
Improve one paragraph by addressing a specific weakness.
We could then reveal the intended bands and discuss the generated commentary. Where students disagree, they would need to explain their reasoning using the criteria. That discussion would be part of the learning.
The strongest response could provide a point of comparison, while the other two could help students identify a manageable next step. “Improve your terminology” becomes something more concrete: distinguish the process from the structure, explain the effect on the relevant lipoprotein, or connect the point to the question.
The generated document labels material as an “official CCEA mark scheme” and an “examiner report”, although those sections are AI-generated. Those labels need removed or changing. I removed the “CCEA mark scheme” label and changed “examiner report” to “marking feedback”.
A polished layout cannot establish that the content or suggested mark is correct. Before sharing the materials with students, I would check the table, grid, essays and commentary against the appropriate CCEA sources.
This builds on the terminology work I already do during lessons, before topic tests and when reviewing answers afterwards. The essay comparison would add an opportunity for students to examine how precision and explanation work within a complete response, then practise improving one themselves.
I have generated the resources, but I am describing the three-band comparison here as a possible classroom use rather than a tested intervention. I cannot yet claim that this activity improves attainment. What I can see is its potential to make feedback more specific and give students something purposeful to do with it.
For teachers of other subjects, the same approach could begin with their own examiner reports, specification and assessment criteria. The question to ask is:
Where does familiar knowledge need a more precise, developed or applied explanation at this level?
AI can help us find the patterns, organise the evidence and generate examples to work with. But it does not decide what an A-level answer should look like: the specification, assessment criteria, mark schemes and examiner reports do that.
The teacher’s role is to bring those sources together and make the step from “what I know” to “how I need to express and apply it at this level” visible to students.
