AQA 7042 · C3.27 - Unacceptable AI use in AO1, AO2 and AO3
Changing AI-generated words does not necessarily make the historical reasoning your own. Learn to recognise when a tool has taken over AO1, AO2 or AO3 achievement, and distinguish that problem from incomplete acknowledgement.
Ownership across the assessment objectives
The Historical Investigation is designed to show what you can do as a historian. The final submission must therefore be the product of your own independent work and thought. Using an AI tool is not automatically the same as misuse, but the tool must not perform the achievement for which you are being assessed.
| Assessment objective | What must remain the learner's own |
|---|---|
| AO1 | Selecting and organising historical knowledge; analysing and evaluating features and relationships; making substantiated judgements |
| AO2 | Analysing and evaluating primary or contemporary source material in its historical context |
| AO3 | Analysing and evaluating different historical interpretations in relation to their historical context |
This means that the important question is not simply, “Did an AI tool appear somewhere in the process?” It is, “Who made the historical choices and demonstrated the knowledge, analysis, evaluation and judgement?” If the tool supplied those assessed decisions or that reasoning, the learner has not independently demonstrated the relevant objective.
AI misuse
AI misuse occurs when a learner uses one or more AI tools without appropriate acknowledgement and/or submits assessed work that is not their own.
Acknowledgement and ownership are connected, but they are not interchangeable. A record of AI use can make an allowed use transparent; it cannot turn AI-produced historical thinking into the learner's own achievement.
Selection and evaluation cannot be outsourced
For AO2 and AO3, choosing evidence is part of the historical work. AQA identifies as unacceptable the use of AI to select and provide evaluation of the primary sources and historical interpretations used in the investigation. The problem remains even if every title exists and every reference is formatted correctly: the learner has delegated the decisions about which material matters and how it should be judged.
There is an important boundary between an unverified lead and an assessed decision. If the centre permits it, preliminary research might produce possible search terms or references. The learner must then find and read the real source or historian's work, verify the reference, decide independently whether it belongs in the investigation, and carry out the AO2 or AO3 evaluation. Asking AI to rank a shortlist, choose the “best” material, explain a source's value or compare historians' arguments crosses that boundary.
Fictional diagnostic case — Lena. Lena is working on a fictional practice investigation about changing public-health policy. She asks a chatbot to choose three primary sources and two historians from a long list, then to supply the value judgement for each source and a comparison of the historians. Lena checks that the publications exist and cites them, but follows the chatbot's selection and evaluations. This is unacceptable: checking the references does not make the AO2 and AO3 choices or reasoning her own.
Copying, paraphrasing and refining
Changing the surface of AI-generated content does not necessarily change who produced the intellectual work. AQA and JCQ identify copying or paraphrasing AI-generated sections and whole responses as misuse when the submitted work is no longer the learner's own. They also identify AI completion or refinement as unacceptable when it means that the work no longer reflects the learner's own skills, analysis or evaluation.
| Fictional workflow | Diagnosis | Why |
|---|---|---|
| Omar copies a complete AI response and adds a citation. | Unacceptable | The response is not Omar's independent work; acknowledgement does not make it his achievement. |
| Mei replaces words and rearranges sentences from an AI-generated analytical paragraph. | Unacceptable if the argument and reasoning remain AI-generated | Paraphrase can disguise the wording without transferring ownership of the thinking. |
| Rosa writes a paragraph, then accepts an AI rewrite that adds the causal links and evaluative judgement. | Unacceptable | The refinement supplies assessed AO1 reasoning that Rosa did not independently demonstrate. |
| Ivo uses an AI feature in a word processor and assumes grammar assistance never needs acknowledgement. | The assumption is wrong | The actual assistance must be disclosed under current JCQ guidance; AQA and centre rules still decide what is permitted, and any added assessed reasoning remains a separate ownership problem. |
Continue this lesson.
You’re previewing NovaLearn without an account. Create a free account to read the rest.
Create free account