Citing ResearchGate and citing Academia.edu are the queries this article has to answer — and almost always the wrong question. The right one is: what do you actually have in front of you? A journal article that the platform merely serves? An author manuscript after peer review but before typesetting? A private PDF with no DOI that exists nowhere else? In a thesis, ResearchGate and Academia.edu are almost always the discovery layer, not the citable record. Writing “ResearchGate” as the publisher signals that you missed the difference.
The examples below are made up and only there for the form — a Keller study in a journal, a Meier author manuscript, a Hofmann upload with no record. Take the structure, not the names.
Related cases this piece does not repeat: which version of a paper to prefer is in open access vs. paywall; what a preprint is and when it is enough is in citing preprints; how to tell whether something was actually reviewed is in how to identify peer-reviewed sources. University theses have their own guide to citing dissertations. Reports and working papers belong under grey literature.
The platform is the finder, not the publisher
ResearchGate and Academia.edu are networks. Authors set up a profile, upload PDFs, answer full-text requests, collect followers. That is useful for research: you find a study sitting behind a paywall, or you notice the same group has already shared an accepted version. It is not a publication path.
Three roles students mix up:
- Discovery. Google Scholar, the library, ResearchGate, Academia.edu — places you arrive. The place of discovery does not automatically belong in the reference list.
- Record. Journal, volume, issue, pages, DOI; or the university repository with a handle; or the publisher with an ISBN. That is what someone outside your seminar can still find in five years.
- File. The PDF you read. It may be the version of record, the author accepted manuscript, or an undated scan. The file alone does not tell you what you are citing.
Using ResearchGate as a thesis source therefore almost never means an entry with publisher ResearchGate. It means you reached a text through the platform and now have to reconstruct the record behind it. The same logic applies to Academia.edu — often messier, because the full text sits behind a sign-up wall and the landing page looks like a citable website.
What the platform is not:
- not a journal (even when “Citations” and a journal name sit at the top);
- not a repository in the sense of ZORA, OPUS, or the ETH Research Collection;
- not a peer-review badge. An upload on ResearchGate is as reviewed as the folder on the author’s desktop. Whether the content was peer-reviewed lives in the journal, not on the profile. That check is in how to identify peer-reviewed sources.
The practical consequence: the list shows journal, publisher, or repository. ResearchGate and Academia.edu stay the route by which you found the PDF — the same way “via Google Scholar” does not go in the parentheses.
Version of record, author manuscript, private PDF
Almost every paper exists in several versions. On ResearchGate they often sit side by side, badly labelled. Fail to name the version and you produce misquotations: different page numbers, different wording, sometimes different figures.
- Version of record (VoR). The publisher’s typeset, final version with a stable DOI. The gold standard. If it exists, you cite it — whether you got the file through the university VPN, Unpaywall, or ResearchGate.
- Author accepted manuscript (AAM), often “author manuscript” or postprint. After peer review, before the publisher’s layout. Typical of the green open-access route. The content is usually the same as the VoR; the pagination is not. Citing an author manuscript means marking the status or — once the VoR exists — citing the VoR and keeping the AAM only as a reading copy.
- Author’s original / preprint. Before review. This is not a ResearchGate special case; the rules are in citing preprints. A preprint uploaded to ResearchGate is still a preprint, even if the platform says “Article”.
- Private or members-only PDF. Full-text request, “private full-text”, a file that lives only in your downloads folder. Not findable for a third party. In most fields, not a reference.
The order that survives marking:
- Is there a VoR with a DOI? → Journal article (or chapter, or book), DOI, no ResearchGate.
- Is there only an AAM, but the VoR is on record (journal, DOI, still behind a paywall)? → You may read the AAM; the list gets the VoR record. Page numbers: those of the version you read, and if in doubt pick a passage you can also find in the VoR — or get the VoR.
- Is there really only the manuscript, with no journal? → Set the status (unpublished / accepted manuscript), use a stable address, and be honest: this is grey literature, often too weak for a core claim.
- Is there only a login PDF? → do not cite it.
“Cite the version you read” remains right — it does not mean “paste the download URL”. It means: take the bibliographic data of the version that carries the claim, and the persistent identifier through which someone else can find that same version.
Resolve the DOI first — then type the entry
The classic error is the ResearchGate landing URL in the list: https://www.researchgate.net/publication/123456789_…. The page moves, the upload disappears, the connector swallowed the DOI. APA 7, Harvard, and German footnote style want the record, not the discovery link.
Before you type a line:
- Look for a DOI on the platform. Many RG and Academia pages show it in small type under the title or in the “Citations” box. Not the internal publication ID — that is not a DOI.
- Open doi.org.
https://doi.org/10.…must resolve to the publisher or repository page, not a dead landing page. What DOI, ISBN, and ISSN are doing in a reference is in DOI, ISBN, ISSN. - Exact title in quotes in Crossref, OpenAlex, or Google Scholar. Scholar often glues preprint, AAM, and VoR together — which is why you click through instead of taking the first hit.
- Library catalogue if nothing digital turns up. Some uploads are scanned articles from issues that received a DOI late or never.
If you find the DOI, https://doi.org/… goes at the end of the entry — not researchgate.net, not academia.edu. Harvard sheets often want an access date for shaky online sources; APA 7 drops it once a DOI resolves. Follow the handout.
Whether the identifier resolves and author, title, year match is the same existence check as for any other citation. The order of steps is in how to check your citations.
Example: This is what a reference list with fabricated and mismatched entries looks like in the check. Check a sample citation →
What the citation looks like
Regardless of style, a clean entry does five jobs: author, year, title, host (journal / publisher / repository / unpublished manuscript), plus the record that lets someone find the text. Drop the host and the study reads like a website. Put ResearchGate there when a journal exists, and the entry is wrong — not incomplete.
Invented templates below. Take the form only.
APA 7 — VoR, found via ResearchGate
In the text like any other author–year source:
Digital supervision shifts responsibility into the documentation, not out of it (Keller, 2023, p. 47).
In the reference list — the journal, not the platform:
Keller, S. (2023). Digital supervision in residential care. Journal of Care Studies, 12(3), 41–58. https://doi.org/10.1234/jcs.2023.00000
No “ResearchGate” after the title. No profile URL. The VoR DOI is enough. Downloading the PDF from ResearchGate does not change the entry.
APA 7 — author manuscript, VoR exists
You read the accepted version; the VoR is on record. Standard in DACH theses: cite the VoR. The AAM is your reading copy.
Only if your style sheet explicitly wants the version you read and you are discussing differences does the manuscript stay beside it:
Meier, L. (2024). Shift rotas and documented supervision [Author accepted manuscript]. Social Policy Review. https://doi.org/10.1234/spr.2024.00000
APA 7 puts the status in square brackets after the title. [Author accepted manuscript] or [Accepted manuscript] are the usual forms; do not mix languages in the same list. Once volume, issue, and pages of the VoR are fixed, those belong in the entry, not the manuscript pagination alone.
APA 7 — PDF with no journal and no DOI
Only the leftover case, and only if your department allows unpublished manuscripts:
Hofmann, A. (2024). Digital documentation in home-care nursing [Unpublished manuscript]. https://www.researchgate.net/publication/000000000_digital-documentation-home-care
This is the entry examiners query. It says: there is no record except this platform. For a core claim that is, in most fields, not enough. For a local pointer (“a draft is circulating in this institute”) it can suffice — if the handbook allows it.
Harvard
Harvard variants differ across universities more than APA does. The order stays: author, year, title, journal or status, then the record.
VoR:
Keller, S. (2023) ‘Digital supervision in residential care’, Journal of Care Studies, 12(3), pp. 41–58. Available at: https://doi.org/10.1234/jcs.2023.00000 (Accessed: 22 September 2026).
A German-language Harvard variant looks closer to this:
Keller, S. (2023): Digital supervision in residential care. Journal of Care Studies, 12(3), pp. 41–58. https://doi.org/10.1234/jcs.2023.00000
Author manuscript, only if the VoR is missing or you are flagging the version on purpose:
Meier, L. (2024) ‘Shift rotas and documented supervision’ [Author accepted manuscript]. Social Policy Review. Available at: https://doi.org/10.1234/spr.2024.00000 (Accessed: 22 September 2026).
Not: ResearchGate in the publisher slot. Many Harvard sheets want an access date for online sources; with a DOI it is redundant unless your handbook insists.
The detail rules for brackets, colons, and available at are in the Harvard referencing guide. APA sits in APA 7. Which system your department wants is in the citation-styles comparison.
German footnote style (Deutsche Zitierweise)
In history, law, theology, and much of the humanities in the DACH region, the citation lives in the footnote. The first mention is a full reference. ResearchGate does not appear there as a publisher.
First footnote, VoR:
¹ Sofia Keller: Digital supervision in residential care, in: Journal of Care Studies 12 (2023), pp. 41–58, here p. 47.
With DOI, if your handbook wants online records in the footnote:
¹ Sofia Keller: Digital supervision in residential care, in: Journal of Care Studies 12 (2023), pp. 41–58, https://doi.org/10.1234/jcs.2023.00000, here p. 47.
Short form later:
² Keller: Digital supervision in residential care, p. 52.
In the bibliography:
Keller, Sofia: Digital supervision in residential care, in: Journal of Care Studies 12 (2023), pp. 41–58.
Unpublished manuscript — only with permission, and then said so:
Hofmann, Anna: Digital documentation in home-care nursing, unpublished manuscript, 2024, https://www.researchgate.net/publication/000000000_digital-documentation-home-care.
A university thesis someone mirrored on Academia.edu stays a dissertation: university, year, repository or DOI — not “Academia.edu”. The pattern is in citing dissertations.
When not to cite a ResearchGate-only upload
The platform makes it easy to treat every file as a source. Four cases where you do not — or at least look for another version.
Nobody but you can get the text. Full-text request, private upload, Academia paywall, a file from a chat. Findability is the minimum bar for a source. What lives only in your downloads folder is not a reference.
The VoR exists, and you pretend the AAM is the journal. Different pages, different abstracts, occasionally corrected figures after review. Anyone who cites Keller 2023 as a journal article but gives manuscript pagination with no volume or issue will be caught in a spot-check.
Unclear origin. Uploads with no author, no date, no journal, or a scan someone other than the author deposited. ResearchGate does not check that. A third-party account can mirror other people’s PDFs. Without a record you treat the file as a rumour.
Your handbook excludes unpublished manuscripts. Many style sheets allow preprints if flagged and ban “other internet PDFs”. An RG upload with no journal falls in the second drawer. When in doubt, one line to your supervisor — not a quiet entry with publisher ResearchGate.
What remains if you still need the study: email the author, interlibrary loan, Unpaywall, the university repository. The platform gave you the name. The name is enough to find the record. If you find none, the claim is, in most fields, not load-bearing for a thesis — the same principle as any other piece of grey literature, only without the institutional sender of a WHO or a government agency.
Web pages without a DOI have their own required fields (access date, host); those are in citing internet sources. ResearchGate is the case where you should not need those fields, because a journal or a repository almost always sits behind the file.
Zotero: clean up the landing page
The connector on a ResearchGate or Academia page is the most common reason entries go bad. Zotero reads what the page offers — and the page offers itself.
Typical traps:
- Wrong item type. Webpage or Attachment instead of Journal Article, Conference Paper, or Thesis. APA then formats title and URL like a website; Harvard hangs “ResearchGate” in the publisher slot.
- Publisher = ResearchGate / Academia.edu. Clear the field as soon as a journal or publisher exists. For a real university thesis the institution goes in the University field, not the platform.
- Missing DOI. The publication ID in the URL is not a DOI. Right-click → retrieve metadata by DOI/ISBN once you have verified the DOI on doi.org.
- URL = profile or publication link. Replace it with
https://doi.org/…. You may keep the RG URL in Extra or as a file attachment if you want to remember where you found the PDF. It does not belong in the rendered list. - Date = upload or “last updated” year. Take the VoR publication year or the year on the manuscript, not the day someone uploaded the file.
- Author = profile name or the platform. Some pages put the account holder above the article authors. Check against the first page of the PDF.
The clean sequence that sidesteps the traps:
- Find the DOI and resolve it on doi.org.
- In Zotero use Add item by identifier (the magic wand) — not “Save page”.
- Attach the PDF to that item. Delete the Webpage ghost from the landing page; do not rename it.
- Hold type, journal, pages, and DOI against the first PDF page and the publisher page.
If there really is no identifier and your handbook allows the manuscript: type Manuscript or Preprint, Extra field with status, URL of the findable version. Not Webpage with publisher ResearchGate — every style formats that wrongly in a different way.
More workflow, without repeating this special case, is in Zotero workflow tips.
Pre-submission checklist
Go through the bibliography once looking only at platform traces — not at commas, at publisher and URL.
- No entry has ResearchGate or Academia.edu as publisher unless there is demonstrably no other record — and then the status is unpublished / manuscript.
- Every study that appeared in a journal has journal, volume, pages, DOI. The RG landing URL is gone.
- Author manuscripts are either replaced by the VoR or flagged as accepted manuscript. No silent mix of manuscript pages and journal names.
- PDF-only uploads without a DOI you have searched in Crossref, OpenAlex, and the catalogue. What remains unfindable comes out — or is honestly listed as an unpublished manuscript if the handbook allows it.
- The Zotero type is Journal Article / Thesis / Manuscript, not Webpage. The DOI field is filled whenever a DOI exists.
- Pagination from the printed version or the VoR, not the viewer numbers 1, 2, 3 of the downloaded scan.
Whether the lines exist and hang together is the same check as for any other citation — the order of steps is in how to check your citations. What remains is the substantive question: does the page you cite actually say the sentence your text claims? The bibliography checker is the way in, and Acurio matches the passage against the uploaded full text. Author manuscripts and VoR PDFs diverge in exactly the places you quote.
ResearchGate and Academia.edu are good search routes and bad publishers. Anyone who keeps discovery apart from journal, repository, and private PDF, resolves the DOI first, and does not let the Zotero connector write the landing page has already done the part examiners mark first. What remains is the version you read — and whether it supports what your text claims.