With books and journal articles the world is simple: author, year, title, publisher, page — done. With web sources it gets messy. Websites disappear, content is revised, a social-media post can be gone within three weeks. Anyone who wants to cite web sources needs a few extra rules — and should make sure before submission that the source will still exist in three years.
Which online sources are actually citable?
Not everything on the internet is suitable for academic work. A rough rule of thumb:
- Citable: government and administrative websites (federal offices, cantons, EU, Federal Statistical Office), academic publishers and repositories (SSRN, arXiv, Zenodo), established media outlets (NZZ, FAZ, SRF, Reuters), specialist blogs with clear authorship, studies from established institutions (OECD, WHO, KOF).
- With caution: industry reports, NGO publications, company websites — citable, but label them as the position of the respective organisation.
- Not citable as evidence: anonymous forums, comment sections, AI-generated overviews without source references. Wikipedia is a starting point, not a source — if you need the content to hold up, follow the footnotes and cite the original sources listed there.
Required fields for a web source
Regardless of citation style — these are the fields you should always collect:
- Author or corporate body (for government sources: the agency, not “n.d.”)
- Year / date of publication
- Title of the item
- Platform / medium (e.g. Der Spiegel Online, FSO, YouTube channel XY)
- URL or DOI (DOI preferred — it’s stable)
- Access date — mandatory for web pages, because content changes
- if applicable, version number (Wikipedia, GitHub) or archive link (Wayback Machine, perma.cc)
If a field is missing, don’t write “n.d.” on every line — leave it out. When the author is absent, the corporate body or the title moves into that position.
Source types in detail
Static website / government publication
Government pages are convenient: stable URL, clear institutional authorship.
Federal Statistical Office (2024): Population level and structure — indicators. FSO, Neuchâtel. https://www.bfs.admin.ch/… (accessed 12 May 2026).
Academic PDF (preprint, working paper)
When a DOI is available, always use the DOI instead of the URL — it survives when the repository migrates.
Acemoglu, D. & Restrepo, P. (2023): Tasks, Automation, and the Rise in US Wage Inequality. NBER Working Paper 28920. https://doi.org/10.3386/w28920
YouTube / lecture with timestamp
For video sources, the timestamp belongs in the in-text citation — just like a page number.
Sapolsky, R. (2017): Human Behavioral Biology, Lecture 5. Stanford University, YouTube, 12:30. https://youtu.be/… (accessed 12 May 2026).
In-text: (Sapolsky 2017, 12:30).
Podcast episode
Episode title, series, publication date.
Köppel, R. (Host) (2024): Episode 142: Federalism in Crisis [Audio podcast episode]. In: Weltwoche Daily, 14 March 2024. https://… (accessed 12 May 2026).
Social-media post (X, LinkedIn, Mastodon)
Full post content in square brackets or shortened with an ellipsis, platform, exact date.
Mariana Mazzucato (@MazzucatoM) (2024, 3 April): The “value” of public investment is not just monetary returns — it’s mission orientation [Post]. X. https://x.com/… (accessed 12 May 2026).
Wikipedia
Never use the “current” URL — use the permalink to the specific version you actually read (on any Wikipedia page, go to “View history” → “Permanent link”). This keeps it verifiable. Wikipedia remains background reading, not a source for central claims.
Federalism (2024, version 12 April 2024). In: Wikipedia. https://en.wikipedia.org/w/index.php?title=Federalism&oldid=123xxxxxx (accessed 12 May 2026).
ChatGPT output
In brief: AI output is a citable source only in exceptional cases (for example, when the output itself is the object of study). In those cases, cite with model, date, and prompt. We have a dedicated post on this — the rule of thumb: AI is not your source, it’s your tool.
Datasets (Eurostat, OECD, FSO)
Eurostat (2025): GDP and main components — current prices [Dataset, Table nama_10_gdp]. Eurostat. https://ec.europa.eu/eurostat/… (accessed 12 May 2026).
For datasets, the exact table or dataset ID is mandatory — without it, an examiner cannot trace the figure back to its origin.
Required fields per source type — overview
| Source type | Author | Date | Title | Platform | URL/DOI | Access date | Special field |
|---|---|---|---|---|---|---|---|
| Government website | yes | yes | yes | yes | URL | yes | — |
| Academic PDF / preprint | yes | yes | yes | yes | DOI | optional with DOI | Working-paper no. |
| YouTube / lecture | yes | yes | yes | YouTube + channel | URL | yes | Timestamp |
| Podcast | yes | yes | yes | Podcast name | URL | yes | Episode number |
| Social-media post | yes (handle) | yes (exact day) | Post text shortened | Platform | URL | yes | Post type |
| Wikipedia | “Wikipedia” | Version date | Article title | Wikipedia | Permalink | yes | oldid |
| ChatGPT | OpenAI / model | Date | Prompt | ChatGPT | — | yes | Model version |
| Dataset | Corporate body | yes | Dataset name | Platform | URL | yes | Table ID |
Archiving — so your source still exists in three years
Websites die. If you write a thesis in mid-2026 and defend it in 2028, you may find that a key government page has been restructured and the link goes nowhere. Two tools to guard against that:
- Wayback Machine (archive.org): when you access a source, click “Save Page Now”. The archived permalink (
https://web.archive.org/web/2026.../...) is stable and can go into the footnote as a second link. - perma.cc: available for free at many universities. Generates short, permanent links specifically for academic work — and archives HTML and PDF at the same time.
In practice: for every online source you cite, create a Wayback snapshot right after saving it. Store the archive link in Zotero’s “Extra” or “Notes” field — if needed, it moves into the footnote.
Common mistakes when citing web sources
- Bare URL without context. “Source: https://bfs.admin.ch/xyz” is not a reference — it’s a bookmark. Author, title, and date must accompany it.
- No access date. Mandatory for web pages. Without an access date there is no way to verify which version you read.
- “Ibid.” for web sources. Works in some styles for books, but is confusing for web sources because the location cannot be pinpointed unambiguously. Better to cite the source in full again, or use a permalink or timestamp.
- Shortened tracking URLs.
bit.ly/xyzor affiliate URLs have no place in academic work — they can be redirected or deactivated at any time. Always use the original URL. - PDF URL vs. landing page. If the agency provides a dedicated landing page with metadata for the PDF, that page is usually the more stable choice. Only cite the PDF URL directly when no landing page exists.
- Auto-import without manual check. The Zotero Connector often pulls “n.d.” as the author and today’s date as the publication date for web pages. Check every web source’s fields manually after import.
Sample check: One click opens a prefilled bibliography and marks fabricated and mismatched entries. Check a sample citation →
Acurio checks your web sources too
Web sources — government reports, long PDFs, Eurostat data — are exactly where a claim is most likely to slip. You take a figure from a 60-page federal PDF; three weeks later you remember it as “four percent” — but the document says “four percentage points.” No plagiarism checker catches that; a second reader very often does.
Acurio reads your sources alongside your text, compares every claim against the original, and flags where source and assertion diverge. Especially for PDFs and government reports, it’s the fastest way to guard against false citations in the final stretch before submission.
Citing web sources is not rocket science, but it forgives fewer errors than citing books. Collect author, date, platform, URL, and access date consistently, use archive.org as a reflex, and work with permalinks on Wikipedia — and you’ve cleared the hurdle. The rest is a clean final review.