Five Common Mistakes to Avoid When Using AI for UK University Assignments

Type a question into a chat window and you get a tidy, essay-shaped answer in seconds. That ease is exactly why so many students come unstuck. The problem is rarely dramatic cheating; it is the small, boring details — a citation nobody can find, a paragraph you cannot explain, a file uploaded somewhere it should never have been.

What follows are the five mistakes that surface most often in marking, feedback sessions and academic integrity conversations, and what to do instead. None of them mean you have to abandon the tools. All of them mean you have to stay firmly in charge of the work.

Mistake 1: Citing sources you have never actually opened

A confident sentence with a neat Harvard citation and a working DOI looks authoritative. It can also be invented. These tools are skilled at producing reference lists in the shape of real scholarship: a plausible author, a real journal, a year that fits, a title that sounds exactly like something someone would write. Sometimes the source exists but says the opposite of what the sentence claims.

Markers have seen it all before. A reference that cannot be found in the library catalogue is a fast route to an awkward email, and "the tool made it up" is not a defence you want to be relying on in a misconduct meeting.

How to check a source properly

1. Search the exact title in your university library discovery service. If it is a journal article, check that the DOI resolves to the right paper, volume and page range.
2. Open the abstract and skim the section relevant to your point. Confirm the claim is genuinely there.
3. Note the page number you are citing. Page numbers you have never seen are a giveaway.
4. Save a copy or a screenshot of the first page, and file it with your notes as you go.
5. If you cannot find it, replace it. One real source you have read beats five impressive ones you have not.

One rule covers most of this: never cite anything you have not read with your own eyes.

Mistake 2: Reading a clean similarity report as permission

A similarity score is not a pass mark and not a permission slip. It measures matching text. It does not tell you whether the rules of your module allow a particular tool, or how much of your draft you were supposed to write yourself.

Policies vary across institutions and often between modules. One module may ban these tools outright in assessed work; another may permit them for planning, explaining a concept or tidying grammar; another may allow almost anything provided you declare what you did. The only version that matters is the one in your module handbook, assessment brief or programme regulations.

If those documents are silent, ask your tutor before you submit rather than after. A short email — "I used a tool to help me structure my draft; is that acceptable under this brief?" — takes two minutes and removes the risk entirely. Detection software also flags human writing from time to time, so a clean result proves less than students assume. Your marker's judgement, not a percentage, is what you are actually working with.

Mistake 3: Pasting in data that is not yours to share

Anything you type into a public tool may leave your control. Depending on the service and your account settings, text can be stored, reviewed or used to improve the model.

That matters whenever your assignment involves real people or unpublished material. Interview transcripts, questionnaire responses, placement notes, pupil or patient information, a supervisor's draft chapter, commercially sensitive documents, another student's essay — none of this belongs in a public chat window. Under UK data protection rules, personal data needs a lawful basis and a clear purpose, and "it was convenient" is neither.

Most universities now provide approved tools or written guidance on what can be used with research data, so check there first. If you are unsure whether something counts as personal or confidential, ask your supervisor or your institution's data protection team before you paste, not afterwards.

Mistake 4: Submitting writing you cannot defend

Assessments rarely end when you upload the file. There are seminars, tutorials, lab debriefs, presentations and follow-up questions. A tutor reading your work will ask what you meant by a particular sentence, why you chose one source over another, or how your argument would change if the evidence pointed the other way.

This is where polished but borrowed prose falls apart. If you cannot explain a paragraph in your own words, you cannot defend it in a room. There is also the quieter problem of voice: a sudden jump in vocabulary, mixed British and American spellings, or a fluency that vanishes the moment you talk about the same topic. Markers notice patterns like that.

Two quick tests

• Explain your main argument in three sentences, out loud, without notes. If you stumble, the understanding is not there yet.
• Highlight every claim in the draft and trace it to a source you can point to. Anything untraceable gets cut or rewritten.

Mistake 5: Losing the paper trail

Students who run into difficulty usually cannot show what they did. Keep a simple log as you work, and leave version history switched on for your drafts.

This is not bureaucracy for its own sake. It shows your working if a question ever arises, it makes clear how much of the final text is genuinely yours, and it makes any required declaration straightforward to complete. A few lines per session is enough:

• The date and the tool you used.
• What you asked it to do.
• What you kept, rewrote or rejected, and why.
• Where any fact you went on to use actually came from.

A ten-minute routine before you submit

Run through this each time, and most of the risks above disappear on their own.

1. Every reference opened and verified.
2. Every claim traceable to something you have read.
3. Spelling and terminology consistent throughout — UK spelling if that is your house style.
4. Module rules re-read, and any declaration completed.
5. No personal or confidential data anywhere near a public tool.
6. The whole document read aloud, start to finish.

The standard you are marked against is not whether a tool was involved. It is whether the work is yours and whether you can stand behind it. Get the boring details right, and the rest looks after itself.

Photo: ernestoeslava / Pixabay