Current Students The Learning Centre Resources AI for university study Evaluating AI output

Evaluating AI output

GenAI tools can produce content that appears polished and authoritative, but this is no guarantee of accuracy or reliability. They may generate fabricated statistics, citations that do not exist and inaccurate information. Relying on AI output without critical evaluation puts the quality and integrity of your academic work at risk.

A structured approach to evaluation helps you identify errors and gaps, and ensures that your work meets the required academic standards. Consider four key dimensions: accuracy, relevance, ethical considerations, and human judgement.

Accuracy

Cross-check key claims against reliable sources. Verify that statistics and citations are real and correctly support what the output is claiming.

Relevance

Does the output address what you actually asked? Does it meet the discipline, complexity, and evidence requirements of your task?

Ethical considerations

Check for bias and balanced perspectives. Respect intellectual property and acknowledge the use of AI in your work.

Human judgement

Apply your own knowledge and expertise. You are responsible for the quality and integrity of the final product.


Practical guidance on evaluation

Expand each dimension below for practical guidance on what to look for and how to act on it.

GenAI generates responses by predicting patterns, not retrieving verified facts. Errors and fabricated content can appear in confident, well-written responses.

  • Cross-check key claims against reliable sources such as peer-reviewed articles, clinical guidelines, and textbooks.
  • Verify that cited statistics and references are real and correctly support what the output is claiming.
  • Treat absolute-sounding statements with scepticism.
  • Read more about how to check GenAI output for credibility.

A factually accurate response can still fail to serve your purpose if it does not match your task requirements.

  • Check whether the output actually answers what you asked or has drifted in scope
  • Consider whether it meets your discipline's expectations for evidence, complexity, and argument
  • Ask whether the tone, language, and depth are appropriate for your intended use
  • Check for bias. Check whether the output presents a balanced perspective across diverse viewpoints, cultures, and experiences.
  • Respect intellectual property and academic integrity. Be clear about the role AI played in producing your work and acknowledge its use.
  • Consider the broader implications of the content. Could it cause harm, reinforce stereotypes, or mislead someone if used without modification?

AI output is a starting point, not a finished product. Your thinking must remain central.

  • Use your own expertise and prior knowledge to assess credibility even before verifying externally.
  • Take ownership of the final product. You are the one responsible for using it in your studies.
  • Keep a record of your prompts and outputs in case you need to account for your process
  • Engage with your subject genuinely and ensure that your own thinking remains central to the academic work.