The Examiner Report Genie: From Common Misconceptions to Targeted Intervention
How I use AI to trawl years of examiner reports, uncover recurring misconceptions, build a diagnostic tool and use the results to decide what to target next.
I have always thought of AI as a bit like having Aladdin’s lamp at my fingertips.
Except, thankfully, I am not limited to three wishes.
As long as I have a useful job for it to do, I can keep coming back with another question, another task or another idea I want to explore. The important part is knowing what to ask for.
This post is one that is certain to benefit every single teacher in Northern Ireland and contains valuable insights and actionable steps on how to use AI to trawl through years of examiner reports, identify recurring misconceptions, turn those findings into targeted retrieval and diagnostic questions, and help work out what your students actually needed you to address before the exams.
Examiner reports are a goldmine of information for teachers and students.
Every year, when the latest examiner reports appear, I read through them and think about just how much valuable information is sitting inside them.
There are recurring misconceptions, terminology students repeatedly confuse, areas where responses lack precision, useful examiner comments about stronger answers and clues about the small differences that can separate one level of attainment from another.
At A-level, there are also those telling moments where students are still relying on GCSE-style language when greater precision, depth and subject-specific terminology are expected.
The problem is not the quality of the information.
It is the time required to extract and organise it.
The information is there. The time for teachers isn’t.
Reading one examiner report is manageable.
But once you have eight or nine years of reports, across every topic in a subject, turning all of that feedback into something you can use systematically in the classroom becomes a mammoth task.
I could work through every report manually, highlighting misconceptions, noting terminology issues and sorting examiner comments into the relevant topics.
It would be useful.
It would also take hours.
And that is only one course.
I teach A-level Biology, Nutrition and Food Science, GCSE Biology and Double-Award Biology with different specifications, then Double-Award Chemistry. The scale quickly becomes unrealistic! It exhausts me just thinking about it!!
That was one of the moments when AI began to feel like my own Aladdin’s lamp. I knew the information was there. I knew what I wanted to find. I wanted to do the best for my students.
I just needed a genie to help with some of the searching and sorting.
Building the examiner-report notebook
The practical setup is simple.
I create a separate Gemini Notebook ( NotebookLM notebook) for each specification and add the relevant examiner reports as sources. For me, that might mean one notebook for GCSE Biology, one for Double-Award Biology, one for AS2 Biology and another for AS1 Nutrition and Food Science.
Once those reports are in the notebook, they become a resource I can return to throughout the year rather than a collection of separate documents that I have to keep reopening and searching. Then the AI genie can start granting wishes.
Stage | What I do |
1. Gather | Download several years of examiner reports for the course. |
2. Upload | Add the reports as sources in a dedicated Gemini Notebook notebook. |
3. Ask | Use a focused prompt to extract evidence for one topic or question. |
4. Check | Verify the AI output against the original reports. |
5. Reuse | Return to the same notebook throughout the year as topics are taught. |
My first wish: find the misconceptions
I began with one topic and one straightforward question:
What misconceptions have examiners identified in relation to this topic?
Comments that had previously been scattered across different years and reports could suddenly be brought together.
So I pushed it further.
Could it look across all of the reports?
Could it identify recurring misconceptions?
Could it organise them according to the topics on the specification?
That is when the real potential became obvious.
I was not asking AI to teach the subject for me or decide what my students needed to know.
I was asking it to help me interrogate a collection of authoritative documents far more efficiently than I could do manually.
The prompt:
Analyse the examiner reports I have uploaded for [subject]. Identify all misconceptions mentioned by examiners and organise them according to the topics in the specification. For each misconception, explain what candidates commonly get wrong, what the correct understanding should be, and identify the examiner report in which it was mentioned. Do not introduce misconceptions that are not supported by the uploaded sources.
That final instruction matters.
I do not want a generic AI-generated list of things students might get wrong. I want the findings grounded in the examiner reports I have deliberately provided.
And, as always, I check the findings against the original source material before using them.
My second wish: turn the misconceptions into targeted retrieval
Initially, I thought the useful part would simply be having a neat list of misconceptions.
But the list is only the starting point.
Once I have checked the findings against the original reports, I can use those same areas of difficulty to create retrieval practice that is deliberately focused on what previous candidates have struggled with.
Targeted flashcards
Rather than creating another generic deck covering an entire topic, I can focus the flashcards specifically on examiner-identified weaknesses.
A card might present:
two commonly confused terms;
an incorrect statement that needs correcting; or
a misconception identified by examiners.
The prompt:
Create flashcards that directly address the misconceptions identified for this topic. Include commonly confused terms and incorrect statements that students need to correct.
The purpose is different from simply generating another set of revision cards.
These are targeted at areas that examiner evidence suggests students have found difficult.
That gives the retrieval practice a much clearer purpose.
Example of Gemini Notebook flash card:
My third wish: create a diagnostic tool for my students
This is the part I find most interesting.
Knowing what previous candidates misunderstood is useful.
But what I really want to know is:
Do the students sitting in front of me misunderstand the same things?
That turns the examiner-report work into a diagnostic tool.
I can use the identified misconceptions to create a multiple-choice quiz. The distractors are particularly important.
Instead of using random wrong answers, I can deliberately build the common misconceptions into the incorrect options.
The prompt:
Create a challenging A-level multiple-choice quiz based specifically on the misconceptions identified for this topic. Use the common incorrect ideas as plausible distractors. Include questions that require students to distinguish between commonly confused terms and apply the correct understanding. Use terminology appropriate for A level.
I can then transfer those questions into Microsoft Forms. If needed, the questions can first be formatted in Word and brought into Forms using Quick Import.
Once the class completes the quiz, Forms gives me item-level response data that I can use to spot patterns. A diagnostic assessment.
Check out my MS Forms Diagnostic quiz here: AS1 Principles of Nutrition: Fat Misconceptions Quiz (AI ElevatED) – Fill out form
If almost everyone gets a question right, I probably do not need to spend valuable revision time reteaching that point.
If a small number of students are struggling, I can target support.
But if a large proportion of the class chooses the same incorrect option linked to a known examiner misconception, that tells me exactly where I need to intervene before the test.
That is much more useful than revising the whole topic again simply because an assessment is approaching.
Putting it into practice: CCEA A-level Nutrition and Food Science
I am now applying this approach to the topic of Fat in A-level Nutrition and Food Science.
I am using several years of CCEA examiner reports to identify recurring misconceptions, confused terminology and areas where explanations lack precision.
Once I have checked those findings against the original reports, I can turn them into two focused resources: targeted retrieval practice and a diagnostic Microsoft Forms quiz.
The aim is not to cover every fact in the topic again.
It is to find out whether the areas previous candidates found difficult are also causing problems for the students in my classroom.
The Forms data can then inform the revision lesson before the topic test.
Secure knowledge can be left alone.
Small pockets of misunderstanding can be addressed with particular students.
Misconceptions appearing across the class can be retaught more explicitly.
That is the part of this process I am particularly interested in: whether examiner evidence from previous cohorts can help me make better decisions about what my current students need next.
Once the students have completed the diagnostic quiz, I will be able to compare the examiner-identified misconceptions with my own class response data and see what, if anything, that evidence changes about the revision lesson.
For me, that is where this becomes more than an interesting AI workflow.
It becomes information I can act on.
The teacher still holds the lamp
There is an important professional point in all of this.
AI can do some of the labour-intensive searching, sorting and first-draft generation, but I still decide what matters.
I check the outputs against the examiner reports.
I decide which misconceptions are worth foregrounding.
I review the flashcards and quiz questions before students see them.
And, most importantly, I decide what the resulting evidence means for my teaching.
The genie does the heavy lifting. The teacher decides what leaves the lamp.
For me, that is the right balance.
AI is not replacing subject knowledge, examiner literacy or professional judgement. It is helping me make much better use of evidence that was already available to me.
Examiner reports have always contained valuable information. AI has not made that information more important.
What it has changed is my ability to do something with it.
Instead of reading an examiner comment, thinking that is useful and then moving on, I can build that feedback into the way I teach, assess and revisit individual topics.
The reports identify the potential problem.
Gemini Notebook helps me uncover and organise it.
Targeted retrieval gives students a chance to revisit and correct their understanding.
Microsoft Forms helps me find out whether the misconception exists in my own classroom.
And my teaching can then respond to that evidence.
That is the point where AI stops simply saving me time.
It starts helping me make better-informed decisions about what I teach next.
The examiner reports were always full of gold.
AI has simply given me a much better way to mine it.
