Referencing and Citation Management

Which referencing style should I use?

The one your institution requires for this assessment, which you should ask about rather than infer. Programmes within a single department frequently differ. Once you know the style, follow your institution's own guide over any general account of it, including published style manuals: local guides routinely diverge on author lists, punctuation and electronic source formats, and markers apply the local version.

How close is too close when paraphrasing?

If any phrase of more than about three consecutive words matches the original, or if your sentence order mirrors the original's, it is too close, and a citation does not fix it. The reliable method is procedural: read until you understand the passage, close the source, write it from your own understanding, then reopen and compare. Technical terms should stay as they are; replacing them with near-synonyms introduces error.

Do I need a citation if I paraphrase rather than quote?

Yes. Quotation, paraphrase and summary all require a citation. The belief that only direct quotation needs one is the single most common misconception about plagiarism, and it produces uncited paraphrase, which is plagiarism.

When do I need a page number?

Always for a direct quotation. In practice also for any specific claim: a finding, a figure, a definition, a particular argument. Not when you are citing a whole work for its overall position. The test is whether a reader could find what you are pointing at without reading the entire source. For sources without pages, use whatever stable locator exists: a section, a paragraph number, a timestamp.

How do I cite a source I found quoted in another source?

Find and read the original if you possibly can, because the intermediary may have misread or compressed it, and because a reference list built on secondary citation shows how far the reading went. Where the original is genuinely unobtainable, use your style's "cited in" convention and list only what you actually read, or both, exactly as your guide specifies. Never cite the original as though you had read it.

Are AI-generated references real?

Frequently not. The characteristic fabrication has plausible authors who really do publish in the field, a real journal, a plausible year, and a title that sounds exactly like something that would exist, and no such article. It is detected by the simplest possible test: trying to open it. Never place a reference in a submitted document that you have not personally retrieved and opened, and never accept an assurance that a reference "appears genuine".

How do I check my reference list before submitting?

Two passes. Forward: every in-text citation has a matching list entry, with the same author spelling and year. Backward: every list entry is cited at least once in the text. Then check ordering and consistency, and spot-check entries against the actual documents, because reference managers import publisher metadata that is often wrong and cannot detect a reference that does not correspond to a real document.

Is it better to have more references?

No. A shorter list where every entry has been opened and says what you claim it says is stronger and safer work than a longer one containing entries nobody has checked. Markers increasingly open references as routine.