Academic Referencing Styles Compared: APA 7th, IEEE, Harvard, and Chicago for Theses

Four systems account for most of what an Indian university is likely to require. Two run on a single published manual, one is a family of institutional variants with no central authority, and one runs two different systems under the same name. Find out which your department has adopted before formatting a single entry, because that document outranks every general guide including this one.

Thesis and article writing format and styles of referencing diverge in a way most guides skip: a thesis follows your institution's handbook for its whole life, while an article follows whichever journal accepts it. Convert at the end, not as you draft.

1. What actually decides your reference style

Harvard is not a manual. The University of York's library states it plainly: the style "originated at Harvard University," has "been adapted by individual institutions," and "there is no set manual or formatting rules, so it is extremely important to check and follow your department's specific regulations." York then proves the point against itself. Its Archaeology department is listed as requiring (Lee 2012, 236), dropping both the comma after the surname and the p. before the page number that the university's main guide uses. One university, two Harvards.

So "we use Harvard" tells you almost nothing, and a supervisor can mark your references wrong when you followed a Harvard guide faithfully. You followed a different one.

APA, IEEE and Chicago each have a governing document you can point at, so disagreements get settled by looking something up. Chicago carries its own complication: it runs two systems, notes-bibliography and author-date, which produce visibly different pages. Saying "Chicago style" without naming the system repeats the Harvard mistake.

One authority sits above all four. Journal house styles override general manuals on submission. Nature requires that "all authors should be included in reference lists unless there are more than five, in which case only the first author should be given, followed by 'et al.'" That cutoff matches none of the systems below. When your handbook and your target journal disagree, they are not in conflict; they govern two different documents.

2. The four-style checklist: book, journal, conference paper, webpage

One format per cell, taken from each system's own published guidance. Harvard appears in the University of York variant, labelled as such, because no canonical version exists.

That last cell is not an oversight. Chicago's freely published Citation Quick Guide runs from books to podcasts without a conference paper entry, so anyone working from the free resource is citing one from memory.

Chicago's two systems are easiest to tell apart on a journal article. Notes-bibliography gives Kwon, Hyeyoung. "Inclusion Work: Children of Immigrants Claiming Membership in Everyday Life." *American Journal of Sociology* 127, no. 6 (2022): 1818–59. Author-date moves the year forward and compresses the issue: Kwon, Hyeyoung. 2022. "Inclusion Work: Children of Immigrants Claiming Membership in Everyday Life." *American Journal of Sociology* 127 (6): 1818–59. Mixing them inside one bibliography is the tell that a thesis was assembled from several guides.

Four further traps come up again and again in formatting reviews.

APA requires the issue number on every journal article, attached to the volume with no space before the parenthesis. Writers trained on the sixth edition still drop it when a journal paginates continuously.

Publisher location has split the systems apart. APA 7 instructs authors not to include it, and Chicago dropped the requirement in its 18th edition, noting that "a place of publication is no longer required in book citations." Harvard as York teaches it still keeps London: Pearson. A thesis that mixes the two conventions looks careless even when everything else is right.

IEEE separates two things students treat as one. A paper you heard delivered is cited as presented at the ... Conf. with no page numbers; a paper in the proceedings is cited in Proc. ... with page numbers, and the guide states that all published conference or proceedings papers have page numbers. If you cannot find a page range, you are probably citing the talk rather than the paper.

Article numbers have replaced page ranges across much of open-access publishing. APA writes the word Article and the number in place of the range, as in *PLoS ONE*, *13*(3), Article e0193972. IEEE writes Art. no. 061103 after the month. Leaving the field blank is the error.

3. Citing non-traditional sources: AI tools, GitHub repos, and datasets in 2026

This is where guides age fastest, and where a supervisor is most likely to be working from something stale.

APA replaced its generative AI guidance entirely. The widely circulated 2023 post on citing ChatGPT now carries a banner saying updated guidance was published in September 2025 and that the post is "retained for archival purposes only." The current reference examples page, created in February 2026, splits the question in two.

For the tool itself, the author is the company and the date is the year it was most recently updated:

OpenAI. (2025). ChatGPT [Large language model]. https://chatgpt.com

For a particular conversation, which the old guidance called uncitable because chats were not retrievable, APA now expects a full date, a chat title in sentence case, a bracketed description, the tool name and a shareable link:

OpenAI. (2026, February 10). Top 10 writing topics for high school graduates [Generative AI chat]. ChatGPT. https://chatgpt.com/share/698b9233-1b58-8008-858d-cede516f4240

The practical consequence is a habit rather than a format. Rename the conversation to something descriptive and generate the share link while you are still in it. Both are far easier in the moment than weeks later. APA also gives the reasoning behind the author element: a model cannot be an author "because it is not a living, conscious human who can give consent and promise to abide by the rights and responsibilities that come with authorship."

Engineering students should know the comparison is uneven. The IEEE Reference Guide in general circulation is version 11.12.2018 and contains no AI entry at all. Where your manual is silent, ask your supervisor and document what you agree, rather than borrowing APA's format and calling it IEEE.

Datasets are settled in both systems. APA places the identifier and version in parentheses after an italicised title, then the repository:

O'Donohue, W. (2017). Content analysis of undergraduate psychology textbooks (ICPSR 36966; Version V1) [Data set]. ICPSR. https://doi.org/10.3886/ICPSR36966.v1

IEEE orders the same information as Title, Source, Date, doi: xxx. Both expect the version where one exists, and it is the element most often dropped even though it is what makes the citation reproducible.

Code repositories are the genuinely unresolved case. APA's public reference examples include no GitHub entry. What APA does publish is a principle you can reason from: where you are citing existing data or statistics, it says to cite the publication in which those data appeared rather than the data set itself, and for a toolbox it says to cite the work in which the toolbox was published. By that logic, a repository described in a published paper is cited through the paper.

For a repository with no accompanying publication, a bare GitHub URL is a poor target because the content behind it changes. GitHub's documentation describes the fix: connect the repository to Zenodo, which "archives your repository and issues a new DOI each time you create a new GitHub release." Cite that DOI and that release and the reference resolves to the exact code you ran. The repository must be public for Zenodo to reach it, and GitHub advises including a licence.

Preprints are simpler than their reputation suggests. APA treats the repository the way it treats a journal, with no italics on the repository name, and the same pattern covers arXiv and bioRxiv:

Hampton, S., Rabagliati, H., Sorace, A., & Fletcher-Watson, S. (2017). Autism and bilingualism: A qualitative interview study of parents' perspectives and experiences. PsyArXiv. https://doi.org/10.31234/osf.io/76xfs

Nature accepts preprints on recognised servers, so the caution is version control rather than legitimacy: if the paper has since appeared, cite the published version.

4. In-text citation placement in multi-author medical and engineering studies

Four systems, four thresholds for when a list of names collapses into et al., and the threshold in the text is not the one in the reference list.

APA's change from the sixth edition catches experienced writers. There is no longer a first citation that lists everyone: a six-author paper is (Martin et al., 2020) the first time and every time after.

That abbreviation creates a collision APA anticipates and few guides pass on. When two different works in one thesis would both shorten to the same et al. form, you spell out as many surnames as are needed to tell them apart. A medical thesis citing two 2021 papers from overlapping consortia is the usual way to meet it. Related exceptions follow the same logic: identical author and year gets letters appended to the date, and first authors who share a surname but differ by initials keep their initials in the text.

Placement differs physically between the systems, which matters when converting a chapter. IEEE numerals sit on the line, in square brackets, inside the sentence punctuation, and the guide treats them as footnote numbers grammatically, so as shown by Brown [4], [5] and as mentioned earlier [2], [4]–[7], [9] are both correct. They can also serve as nouns: as demonstrated in [3].

Pointing at part of a source is where IEEE is least like the author-date systems. Instead of a page number after a comma, IEEE extends the bracket: [3, Th. 1], [3, Lemma 2], [3, pp. 5–10], [3, eq. (2)], [3, Fig. 1], [3, Sec. 4.5]. A reader can be sent to a specific theorem rather than a page that happens to contain it.

Two conventions converge. Neither APA nor IEEE permits ibid.: APA requires author and year in every parenthetical citation, and IEEE instructs editors to eliminate ibid. and op. cit. and repeat the earlier number. Chicago is the exception, using shortened notes such as Yu, *Interior Chinatown*, 48.

One check is worth running mechanically before submission. APA states the rule directly: each in-text citation must correspond to only one reference list entry. The reverse holds as a matter of practice in all four systems, since a reference list is a list of what you cited. Renumbering an IEEE list after inserting a source is the classic way to break both halves, which is why the guide warns that editing references entails careful renumbering of the list and the citations together.

5. Where Sovi.AI fits

A caution belongs first, and it comes from the organisation that publishes one of these four systems.

When APA's style team tested generative AI in 2023 and asked for sources, it returned five references. Four could be found. The fifth, in the team's own account, "does not seem to be a real article; the digital object identifier given for that reference belongs to a different article." The guidance in that post has been superseded; the finding has not. Never ask any model to produce a reference, a DOI or a page number. Those come from the source in your hand.

That leaves a narrower job that is real. Sovi.AI Smart Writing carries an Academic preset among its tone options, takes file attachments, and offers a second step where you feed it earlier writing of your own so the register comes back closer to yours. Its use here is the prose around the citations, not the citations: tightening a literature review that has turned into a list of surnames. Results appear in an editable panel with a copy button and nothing else, so moving them into your manuscript template is a manual step.

What it does not do should be said directly. Smart Writing has no citation checker, no bibliography generator and no style conversion; nothing on the page reads a reference list or knows what APA or IEEE is. Converting between styles stays a manual check against the kit below, or a job for the reference management software your department supports. Which structural conventions change when a thesis chapter becomes a paper is covered in the guide to thesis format versus journal articles.

Frequently Asked Questions

Q1: My department says "Harvard" and nothing else. What do I do?

Ask which guide they mean and get the answer in writing, since Harvard has no governing manual and institutional variants differ in punctuation, the use of pp., and the author cutoff for et al. If nobody can name one, pick a published version, state in your front matter which you followed, and apply it without exception.

Q2: Can I cite an AI chat conversation in my thesis?

Under APA's current guidance you can, and the format expects a specific date, a descriptive chat title and a shareable link. Whether you may use the tool at all is a separate question governed by your institution's academic integrity policy.

Q3: How do I cite a GitHub repository properly?

If a published paper describes the code, cite the paper. If not, archive a release through Zenodo, which issues a DOI for each release, and cite that DOI with its version. A bare repository URL points at code that may have changed since you used it.

Q4: Does IEEE really want et al. after only three authors?

In the text, yes. The IEEE Reference Guide directs authors to use et al. when three or more names are given for a reference cited in the text. The reference list is more generous, naming all authors up to six before abbreviating.

Q5: When do I convert my references to the journal's style?

After the text has stopped changing. Converting early means doing it twice, and each pass is a chance to introduce errors in the elements that differ between systems: publisher location, issue numbers and the author cutoff.

Q6: Which Chicago system should a thesis use?

Whichever your department specifies, since notes-bibliography and author-date are not interchangeable and produce different pages. Humanities departments typically require the first and science departments the second, but the handbook decides, not the discipline.