Experiment #1: Can AI Save Me Time Creating Recall Quizzes?
When Ai(my only experience so far is Co-Pilot and more recently ChatGPT) first appeared, like many teachers, I was curious.
Not because I thought it was going to revolutionise education overnight, but because I wanted to know whether it could genuinely help with some of the more repetitive jobs that come with teaching.
Just after Christmas last year, I knew I had a much bigger project ahead of me. Our KS3 Computer Science curriculum needed a new Ai unit and I wanted to create a completely new Scheme of Learning. Before trusting AI with something as important as curriculum design, I decided to start much smaller.
Could it create useful recall quizzes?
Creating retrieval quizzes is one of those jobs that takes longer than people realise. Writing one quiz isn't difficult, but when you're creating one for every lesson across an entire Scheme of Learning, the hours soon add up.
It seemed like the perfect place to begin my experiments.
The Question
Can generative AI create recall quizzes that I'd actually be happy to use in the classroom?
Not just quickly.
Not just in large quantities.
But quizzes that accurately assess the key knowledge from a lesson and help students retrieve what they've learned.
The Experiment
My first attempt was deliberately simple.
I uploaded the PowerPoint from an existing lesson and asked ChatGPT to generate a 10-question multiple-choice recall quiz based on its contents.
The goal was simple.
If this worked, I could potentially create recall quizzes for every lesson in a Scheme of Learning in a fraction of the time it would normally take.
My original prompt looked something like this.
Prompt 1
The results were surprisingly good...but they weren't always consistent.
The Result
The short answer?
Yes...but not consistently.
ChatGPT had no problem producing quizzes at scale. Within seconds I had a complete recall quiz.
The challenge wasn't speed.
It was quality.
Some quizzes were excellent and tested exactly what I wanted students to remember.
Others focused on minor details, occasionally missed important concepts or simply didn't have enough information to work from.
I quickly realised something that seems obvious now.
My teaching PowerPoints weren't written as standalone learning resources.
They were designed to support my explanations during a lesson.
AI only sees what's written on the slides.
It doesn't hear everything you say.
What I Changed
Like most things with AI, better input produced better output.
Instead of simply asking ChatGPT to create a recall quiz, I became much more specific about what I wanted.
After several iterations my prompt evolved to include instructions such as:
-
Generate 10 multiple-choice questions
-
Questions should become progressively more challenging
-
Include five answer options
-
Randomise the position of the correct answer
-
Produce an answer key
That last point might sound trivial, but one of my earliest quizzes had a worrying habit of making Option B the correct answer far too often!
Here's the updated version of the prompt.
Prompt 2
I also realised that the quality of the source material mattered just as much as the prompt.
Where possible I began uploading:
-
OCR specification statements
-
lesson notes
-
transcripts from YouTube videos
-
textbook extracts
-
additional teacher notes
The richer the information, the better the quizzes became.
For A Level topics, I found that uploading the specification alongside the lesson PowerPoint produced particularly reliable results.
Example Output
One of the examples I tested was from OCR A Level Computer Science – Systems Software (1.2.1 a–d).
The resulting quiz looked like this.
A-Level Computer Science Recall Quiz
Systems Software (1.2.1 a–d)
Question 1
What is the primary purpose of an operating system?
A. To permanently store files
B. To create application software
C. To manage the computer's hardware and software resources
D. To connect computers to the internet
E. To increase the speed of the CPU
Question 2
Which memory management technique divides memory into fixed-size blocks called pages?
A. Paging
B. Segmentation
C. Virtual Memory
D. Caching
E. Defragmentation
Question 3
What is an interrupt?
A. A type of CPU register
B. A section of RAM
C. A scheduling algorithm
D. A program stored in the BIOS
E. A signal that requests the CPU's attention
Question 4
Which scheduling algorithm always executes the process that arrived first?
A. Round Robin
B. First Come First Served (FCFS)
C. Shortest Job First
D. Shortest Remaining Time
E. Multi-Level Feedback Queue
Question 5
Why is virtual memory used?
A. To make the processor run faster
B. To permanently store application software
C. To reduce the size of RAM modules
D. To allow programs to continue running when RAM is full by using secondary storage
E. To improve internet performance
Question 6
What happens immediately after the CPU receives an interrupt?
A. The CPU finishes executing its current instruction before responding
B. The computer immediately shuts down
C. The current program is deleted
D. The interrupt is ignored until the program ends
E. The BIOS restarts automatically
Question 7
Which scheduling algorithm gives every process a fixed amount of CPU time before moving to the next process?
A. First Come First Served
B. Shortest Remaining Time
C. Round Robin
D. Multi-Level Feedback Queue
E. Priority Scheduling
Question 8
Which statement best describes segmentation?
A. Memory is divided into equal-sized pages.
B. Secondary storage replaces RAM.
C. Programs are stored alphabetically.
D. Each process is allocated exactly the same amount of memory.
E. Memory is divided into variable-sized sections based on logical program components.
Question 9
An operating system is using the Shortest Job First scheduling algorithm. Four processes are waiting with estimated execution times of 8 ms, 3 ms, 6 ms and 2 ms.
Which process will run next?
A. 8 ms
B. 2 ms
C. 3 ms
D. 6 ms
E. The one that arrived first
Question 10
A user is editing a document when they press a key on the keyboard.
Which sequence correctly describes what happens?
A. The ISR runs → Interrupt occurs → CPU saves state → Program resumes
B. CPU saves state → Program resumes → Interrupt occurs → ISR runs
C. Interrupt occurs → Program resumes → ISR runs → CPU saves state
D. Interrupt occurs → CPU completes the current instruction → CPU saves its state → ISR executes → Original program resumes
E. Interrupt occurs → Computer restarts → ISR executes → Program resumes
Answer Key
| Q | Answer |
|---|
| 1 | C |
| 2 | A |
| 3 | E |
| 4 | B |
| 5 | D |
| 6 | A |
| 7 | C |
| 8 | E |
| 9 | B |
| 10 | D |
What I Learned
This experiment taught me something that has stayed with me throughout every AI experiment since.
AI isn't really about writing better content.
It's about giving it better information.
If your PowerPoint contains very little text because you provide the explanation during the lesson, don't expect AI to magically fill in the gaps.
Instead, provide richer sources of information such as:
-
specification statements
-
lesson notes
-
textbook extracts
-
transcripts from videos
-
model answers
The better the context, the better the output.
I also discovered that prompt engineering genuinely matters.
Small additions—such as asking for progressively harder questions, believable distractors and randomised answer positions—made a noticeable difference to the finished quiz.
Would I Use It Again?
Absolutely.
In fact, it's something I now use regularly.
Once I'd refined both the prompt and the information I was providing, AI became an excellent first draft generator.
Would I use the very first quiz it produced without checking it?
No.
Would I happily let AI produce a first draft that I could review in a couple of minutes?
Every time.
For me, this is one of the first classroom tasks where AI genuinely saved me time without compromising quality.
It doesn't replace teacher expertise.
It simply gives me more time to use that expertise where it matters most.
Experiment Verdict
Would I Use It Again? Yes.
Key Takeaways
-
AI is excellent at generating recall quizzes when it has enough context.
-
A PowerPoint isn't always enough—adding specifications, lesson notes or video transcripts dramatically improves the results.
-
Small prompt improvements can make a surprisingly big difference.
-
Always review the final quiz before giving it to students.
-
AI works best as a first draft, leaving the teacher to apply their professional judgement.
If you like this blog check out all the others here
