GUEST POST

Effective Use and Pitfalls of AI in A-Level Computer Science

AI is already part of the way many A-Level Computing students study. It can explain a difficult concept in seconds, identify an error in a program, generate revision questions and provide suggestions when a student is stuck.

There is clearly a place for it.

The problem starts when AI moves from supporting learning to replacing it.

As a Computer Science examiner, teacher and tutor, I think this is the distinction students need to understand. A-Level Computer Science is not simply about arriving at a correct answer or producing a working program. Students need to understand the processes involved, apply their knowledge to unfamiliar situations and, particularly in programming, be able to explain why their solution works.

An AI tool can produce an impressive answer very quickly. That does not necessarily mean the student has learnt anything from it.

Using AI to understand difficult topics

One of the best uses of AI is as an additional source of explanation.

Computer Science contains a number of concepts that students can initially find difficult: recursion, abstraction, Boolean algebra, object-oriented programming, data structures and networking are common examples.

If the explanation in a textbook or lesson has not quite made sense, there is nothing wrong with asking AI to explain the concept differently.

The important part is what the student asks it to do.

There is a considerable difference between:

“Explain recursion.”

and:

“I think recursion is when a function calls itself. Ask me questions to check whether I properly understand recursion, including the importance of a base case.”

The second prompt requires the student to participate.

AI can also be useful for generating practice:

  • “Give me five questions on stacks and queues, becoming progressively more difficult.”

  • “Give me a trace-table question but do not provide the solution until I have attempted it.”

  • “Give me a scenario where I have to decide which data structure would be most appropriate.”

  • “Here is my explanation of packet switching. Tell me what is technically inaccurate or missing.”

Used in this way, AI can provide additional opportunities to practise and receive feedback.

It should not, however, become the only source a student relies upon. AI can produce inaccurate information and can present that information very convincingly. The Department for Education specifically identifies inaccurate or false information, bias and misinformation among the risks associated with generative AI (Department for Education, 2025).

For A-Level students, there is another issue: an answer can be technically correct but still not be appropriate to the specification they are studying.

The specification and the awarding body's materials should therefore remain the point of reference.

Programming: probably the most useful area but also the easiest to misuse

In A-Level Programming, AI can be particularly valuable.

It is also where I would exercise the most caution.

If a student asks AI to write a complete solution every time they encounter a problem, they may finish the task more quickly, but they are avoiding the part of programming that actually develops their ability.

Debugging is a good example.

Suppose a binary search fails to find the final value in a list. A student could paste the program into an AI tool and ask it to correct the code. There is a reasonable chance that it will.

A better prompt would be:

“My binary search works for most values, but sometimes fails at the end of the list. Do not correct my code. Give me one hint about what I should investigate.”

That keeps responsibility for solving the problem with the student.

The same principle can be applied elsewhere. AI can be asked to identify possible test cases, explain an error message, compare two possible algorithms or question why a particular approach has been chosen.

My test would be straightforward: after using AI, can the student explain the solution without it?

Can they explain what every important section of the program does? Can they justify the algorithm they selected? Could they modify it if the requirements changed?

If not, there is a gap in understanding, irrespective of whether the program works.

Be careful using AI to mark examination answers

Using AI for examination practice can also be helpful, but I would be particularly cautious about treating it as an examiner.

Students sometimes assume that if an AI tool awards an answer six marks out of six, that is a reliable indication of what would happen in the examination.

It isn't.

Mark schemes are applied in the context of a particular question, and examination marking involves professional judgement. AI can overlook an incorrect statement, infer something the student has not actually written or create marking points that are not in the official mark scheme.

A polished answer is not necessarily a high-mark answer.

For examination practice, I would recommend a simple order:

  1. Answer the question without AI.

  2. Check the official mark scheme.

  3. Identify what you missed.

  4. Use AI to help explain anything you still do not understand.

  5. Rewrite or improve the answer yourself.

From an examiner's perspective, relevance and precision matter. Adding more technical terminology does not automatically improve an answer. Students need to respond to the question that has actually been asked and apply their knowledge to the scenario where required.

This is one area where AI-generated responses can sometimes be misleading. They can sound very convincing while remaining quite generic.

The NEA is different

The area where students need to be most careful is the A-Level Computer Science NEA.

This is not simply another piece of homework.

The project is specifically designed to assess the student's ability to analyse a problem, design and develop a solution, test it and evaluate the outcome. In the AQA specification, for example, 42 of the 75 available project marks are allocated to the technical solution alone (AQA, n.d.).

That is significant. The programming is supposed to demonstrate the student's technical ability.

OCR's guidance on AI use in A-Level Computer Science makes the position equally clear: only independent candidate work should receive credit. OCR recognises that AI can be useful as a support tool but warns against over-reliance and the use of AI for substantial sections of the work (Cattanach-Chell, 2026).

There is therefore an important difference between AI helping a student and AI completing the project for them.

Where its use is permitted, reasonable support might include using AI to:

  • help generate questions that could be asked during research;

  • explain an unfamiliar programming error;

  • identify areas that require further investigation;

  • suggest possible test cases that the student then considers;

  • question the student's reasoning;

  • explain a programming technique that the student subsequently implements independently.

I would be much more concerned if AI were being used to:

  • generate substantial sections of the program;

  • produce the project's analysis;

  • write pseudocode or the detailed design;

  • manufacture stakeholder feedback;

  • create testing retrospectively;

  • write the evaluation;

  • rewrite sections specifically to satisfy assessment criteria.

Changing variable names in AI-generated code does not suddenly make the programming independent work. Equally, paraphrasing AI-generated analysis does not demonstrate that the student carried out the analysis.

The question should always be: what has the student demonstrated independently?

That is ultimately what can be credited.

If AI has been used, do not hide it

Trying to disguise AI use creates a much greater problem.

JCQ guidance requires students to acknowledge AI where it has been used as a source. This includes identifying the tool and the date on which the content was generated. Students should also retain their prompts and the AI responses, and explain how the material was used (JCQ, 2025).

For an NEA, I think keeping a simple record is sensible practice.

Record:

  • when AI was used;

  • which tool was used;

  • what was asked;

  • what response was provided;

  • how the information was checked;

  • what the student subsequently did themselves.

This creates transparency and, more importantly, encourages students to think about the extent to which the tool has influenced their work.

Acknowledging AI does not mean that AI-generated work can then receive marks. JCQ is explicit that students cannot receive credit where they have not independently met the assessment criteria (JCQ, 2025).

That distinction is sometimes missed.

Do not remove the difficult part of learning

There is a wider issue beyond examinations and the NEA.

Students become better programmers by solving problems.

That includes getting things wrong.

Finding an error, trying a solution that does not work, tracing a program, looking at documentation and eventually identifying the problem are all useful parts of learning Computer Science.

If the first response to every difficulty is to ask AI for the answer, those opportunities disappear.

I would therefore suggest the following approach when a student becomes stuck:

  1. Try to solve the problem independently.

  2. Go back to notes, examples or documentation.

  3. Work out exactly what part is causing the difficulty.

  4. If necessary, ask AI for an explanation or a hint rather than the solution.

  5. Apply that information yourself.

  6. Test the result.

  7. Make sure you can explain what you have done.

The purpose of using AI should be to improve understanding. If it allows a student to complete more work while understanding less, it is being used badly.

Final thoughts

I do not think the answer is to tell A-Level Computer Science students simply not to use AI. They are going to encounter these tools at university, in apprenticeships and throughout the technology industry.

Learning how to use them appropriately is therefore becoming a useful skill in itself.

However, students still need the underlying Computer Science knowledge.

They need to be able to program. They need to debug. They need to analyse problems, select appropriate algorithms, interpret questions and justify their decisions.

AI can help with all of those things.

It should not be the reason a student can do them.

References

AQA (n.d.) A-level Computer Science 7517: Non-exam assessment - the computing practical project. Available at: https://www.aqa.org.uk/subjects/computer-science/a-level/computer-science-7517/specification/subject-content/non-exam-assessment-the-computing-practical-project (Accessed: 30 August 2026).

Cattanach-Chell, C. (2026) ‘Guidance and support for the use of AI in A Level Computer Science NEA’, OCR, 4 February. Available at: https://www.ocr.org.uk/blog/guidance-and-support-for-the-use-of-ai-in-a-level-computer-science-nea/ (Accessed: 30 August 2026).

Department for Education (2025) Generative artificial intelligence (AI) in education. Updated 12 August 2025. Available at: https://www.gov.uk/government/publications/generative-artificial-intelligence-in-education/generative-artificial-intelligence-ai-in-education (Accessed: 30 August 2026).

Joint Council for Qualifications (JCQ) (2025) AI use in assessments: Your role in protecting the integrity of qualifications. Effective from 30 April 2025. Available at: https://www.jcq.org.uk/knowledge-hub/ai-use-in-assessments-your-role-in-protecting-the-integrity-of-qualifications/ (Accessed: 30 August 2026).

Author's profile picture

Evan D

Tutor

Highly experienced OCR, AQA, CIE A-Level and GCSE Computer Science Examiner and Teacher. Bespoke OCR and AQA NEA support

Looking for a tutor?

Sherpa has hundreds of qualified and experienced UK tutors who are ready to help you achieve your goals. Search through our tutors and arrange a free 20 minute introduction through our industry-leading online classroom.

Find a Tutor