ÄûÃÊÊÓƵ

Using generative AI in your assessments

What is generative AI?

Generative AI refers to tools that can create new content such as text, images, or code in response to questions or prompts. This page focuses on text-based tools.

Tools like ChatGPT, Microsoft Copilot, Gemini, and Claude are examples of large language models (LLMs). These models are trained on very large collections of text from sources such as books, articles, and webpages. During training, they learn patterns in how language is used so that they can predict the next likely token (a word or part of a word) in a sequence.

Because of this, their responses are generated using probability rather than human-like understanding. This means they can produce text that sounds convincing but may include mistakes, missing context, or incorrect information.

Generative AI tools can produce content quickly and can be helpful for brainstorming or exploring ideas. However, they can also generate inaccurate or misleading information (“hallucinations”). If you rely on AI-generated text in your academic work without checking it carefully, you might receive a poor mark or risk breaching academic integrity expectations.

Always critically evaluate and verify any output before using it.

How do I access Microsoft CoPilot?

Microsoft CoPilot is now available for ÄûÃÊÊÓƵ students. To access, go to the and then sign in with your Sussex username and password.

The University’s subscription doesn’t retain your prompts or responses, and they aren’t used to train the model. All data is encrypted. Please note that Copilot is currently only supported in the Chrome and Edge browsers.

What are the limitations of AI?

• Reliability: AI models can produce well-written and plausible responses, but they frequently make factual errors (called “hallucinations”), fabricate quotes and citations, or present only one side of an argument. This makes them unsuitable for unmediated use in academic work. Increasingly, generative AIs are producing online content, including fake news and deepfakes, which can add further confusion.
• Currency: Generative AI tools do not always have up-to-date knowledge and may be unable to provide reliable information about current events.
• Specialist areas: They perform better in subjects that are widely written about, and less well in niche or specialist areas.
• Bias: Generative AIs are not free from bias. Because they are trained on large amounts of human-produced text, much of it in English and from Western contexts, they can reflect and reproduce harmful stereotypes or cultural biases. There are also concerns that as AI-generated content becomes more common online, newer models may train on this material, leading to reduced quality and diversity of information (“model collapse”).
• Ethics: In some cases, AI models have been trained on copyrighted or unlicensed material without the creator’s consent. Depending on the tool, your interactions may be used to improve the model unless you opt out. Many tools also rely on human reviewers (a process called Reinforcement Learning from Human Feedback) to evaluate outputs, raising ethical concerns about labour conditions for content moderators.
• Access/Equity: The full functionality of AI tools are usually only available via subscription, which we would discourage anyone from doing on an individual basis.
• Privacy: You should never share personal or sensitive information with AI tools. Some platforms allow you to opt out of having your conversations used for training, but this depends on the tool and the type of account.

It’s important to remember that anyone can upload content to the internet, and generative AI models are trained on large datasets that may include inaccurate, biased, or copyrighted material. You should always critically evaluate any source—AI-generated or otherwise. The Skills Hub has many resources on critical thinking, including approaches like the CRAAP test.

Can I use generative AI to edit or proofread my assessments?

You are responsible for ensuring that all assessed work you submit is entirely your own. This includes the structure, argument, analysis, interpretation, and conclusions. Proofreading—whether by a person or an AI tool—can only support you in very limited ways. Any help that goes beyond these limits may be considered academic misconduct, including personation, collusion, or unauthorised use of digital technologies.

A proofreader (person or AI) may not:

Remember: A proofreader may not be used for assessments where the use of language and the formal accuracy of the work form part of the mark.

A proofreader (person or AI) may:

Any suggestions made by a person or an AI proofreading your work should be retained in case of any concerns about academic misconduct. We recommend saving and writing your assignments using your ÄûÃÊÊÓƵ OneDrive account as this will automatically keep version history.

Students registered with the Disability Advice team and in receipt of reasonable adjustments are still permitted to use other assistive technology as recommended by the University in their Learning Support Plan.

Support and guidance available:

• Skills Hub for more guidance on proofreading

Can I use Grammarly?

You are free to use Grammarly’s basic grammar, spelling, punctuation and expression-correction features to improve the clarity of your writing. These features function similarly to a proofreader making minor surface-level edits. However, you must not use Grammarly’s generative AI features (including: drafting new text, rewriting paragraphs, reorganising your argument, paraphrasing large sections, or generating references) unless the assessment instructions explicitly permit such use and you follow any required acknowledgement guidelines.

The same rules apply to Grammarly as to any proofreader or support tool. You can use Grammarly for improving your expression, but not for reshaping your ideas or arguments. Misusing Grammarly’s generative AI features may constitute academic misconduct such as unauthorised use of digital technologies or personation.

Can I use generative AI tools to translate my work?

You should only use translation tools—including AI translation—if your assessment guidance clearly states that they are permitted. It is important to note that minor (single word) translation is permitted, however, translating substantial sections of text is not.

All work you submit must be written by you, in the language required by the assessment. It must not be translated or rewritten by a third party—whether a person or an AI tool.

If a translation service is required for an assessment, you will need to confirm which section(s) of the assessment were translated and whether the translation was done by you or by someone/something else.

If you need support with writing, the University offers a range of support:

• English Language for Academic Study (ELAS) - and for students working in English as an additional language.
• Academic Skills Workshops and Webinars - guidance on writing, structure and academic techniques
• Royal Literary Fund Fellows - 1-2-1 writing support
• Additional writing support may also be available through your School or Department

Can I use generative AI to write my assessments?

You should only use generative AI if your School has explicitly allowed it for a specific module or assessment. If AI use is permitted, this will be stated clearly in your assessment instructions. Where AI use is allowed, you must follow the guidance provided and acknowledge any AI-generated content

Using AI in an assessment where it is not permitted, or submitting AI-generated work without the required acknowledgement, is academic misconduct under the University’s policy on the misuse of digital technologies.

All work submitted for assessment must be your own, and you must fully acknowledge every source you use.

Any action that gives you an unfair advantage in an assessment or exam is considered academic misconduct. For further support, see Skills Hub guidance on referencing and academic integrity.

Can I use AI tools to find references for my essays?

Large Language Models (LLMs), such as ChatGPT, are not databases and do not have direct access to academic sources. Because they generate text based on patterns in their training data, they often invent (hallucinate) references, quotations and citations. Submitting fabricated references is a form of academic misconduct.

This is different from using tools designed to help you search for real research. Platforms such as Elicit or Consensus work by retrieving verifiable academic sources, rather than generating invented ones. If you choose to use such tools, you must still read the sources yourself and check that the details are accurate.

Remember, in academic writing, you should always be led by the literature — read and engage with sources first, then build your argument — rather than writing your essay and trying to find evidence to support it afterwards. Academia is a conversation, and your essay should show how your ideas respond to, build on, or challenge existing scholarship.

How do I know if the use of AI is permitted on my module?

If the use of generative AI is allowed for a particular module or assessment, this will be stated clearly in your assessment guidance. If AI use is not permitted, it is because using it would prevent you from developing the knowledge or skills that the assessment is designed to test. Students with reasonable adjustments recommended by the Disability Advice team may still use assistive technologies included in their Learning Support Plan.

If you are unsure what is allowed, check with your module convenor. Your assessment guidance will contain one of the following statements:

If AI use is permitted, you must keep a record of the prompts you used, the outputs generated, and how those outputs informed your work. You must also follow any instructions on how to acknowledge AI-generated content, as required by the University’s Academic Misconduct Policy.

Remember: Permissions only apply to the specific assessment and module. Do not assume that AI use is allowed in other modules.

Can I use AI tools to support my learning?

Yes, in limited ways, AI tools can assist with your learning.

• Explain difficult concepts or theories in clearer or simpler language
• Summarise texts, transcripts or notes, or present ideas in different formats to help you understand them (e.g., as a diagram, song or story)
• Support your planning, such as suggesting possible structures for an assignment
• Improve grammar and writing style, as long as you remain in full control of the ideas and content
• Help you identify academic literature to explore further
• Debate or test ideas to develop your thinking

Make sure you read your assessment guidance carefully as it will explain exactly how you’re allowed to use AI and what’s not permitted.

AI tools can be useful for building understanding, but you must not copy AI-generated text into your assignments. Submitting AI-generated content as your own work would be considered academic misconduct.

How do I reference the use of AI in my work?

If your module or tutor has given you explicit permission to use AI, you should include an acknowledgement in your work. A suggested format is:

AI tools cannot be cited in the same way as books or journal articles because they do not have stable, traceable sources, and they are not considered authors. AI-generated content is produced from training data and cannot be referenced as an original, retrievable source.

What information should I avoid sharing with AI tools?

Before using any AI tool, make sure you read its terms of use. Many AI systems store the data you provide and may use it to train future models. Any personal or sensitive information you enter could be stored, shared, or even made public.

If you’re working with ÄûÃÊÊÓƵ data, always use . This version is licensed and keeps your data protected. If you use any other AI tool, make sure to opt out of data collection whenever possible.

Even when using Copilot, never share sensitive information such as names, addresses, health data, passwords or unpublished research.

Alternative tools to generative AI

It’s important to understand that there is an ever increasing range of AI tools available and the quality of responses will vary depending on a number of factors including the quality of the prompts, whether you are using the free or subscribed version and if it’s the right tool for the job.