CiteOwl

Blog/Finding and reading sources

How to find peer-reviewed articles without a library login

The CiteOwl Team· Updated 11 August 2026· 19 min read

Three free services do the whole job between them, and none of them asks which university you are at: OpenAlex to find the work, the ISSN Portal and DOAJ to confirm the journal is a real refereed venue rather than a lookalike, and Unpaywall to open the paper. If you do have a library, start there instead and switch on its "peer reviewed" limiter, which is still the least effort for the most filtering. This guide covers both routes, what each index does and does not hold, and the checks that separate a genuinely refereed journal from one that only prints the word.

Peer review is the check that sits in front of serious academic publishing: before a journal publishes a study, it sends the manuscript to other experts who probe the method, the analysis, and whether the conclusions follow from the data. When an assignment says "use peer-reviewed sources," that checked layer of the record is what it is asking for. Two things go wrong. Students cannot find the work, and once they find it they cannot tell a refereed journal from an imitation. The second problem has grown teeth in the last few years, so half of this page is about it.

The free route, start to finish

Find the work in OpenAlex. It is an open index of scholarly works, authors, institutions and journals, and it is the closest free thing to the big subscription citation indexes. Queried on 11 August 2026 it reported 323,998,640 works and 255,343 sources, which is its word for journals and other venues. Use it through the site, or put a query straight in your address bar: https://api.openalex.org/works?search=urban heat island mitigation. One change to know, because most guides still print the old instructions: since 13 February 2026 OpenAlex asks for an API key and the old email parameter is gone. The key is free from your account settings. Without one you get a small daily allowance and then a 429 that tells you plainly what happened.

Confirm the venue. Take the journal's ISSN and look it up at https://portal.issn.org/resource/ISSN/ plus the number. The register is run by the ISSN International Centre in Paris and, as of 11 August 2026, covers over two million serial publications. One thing that trips people up: the print and online editions of one journal are two separate records, so 0028-0836 and 1476-4687 are both correct for Nature. Two ISSNs is normal. None is not.

Open the paper. Unpaywall answers the paywall question for any CrossRef DOI: https://api.unpaywall.org/v2/ plus the DOI plus ?email= and your own address. It returns every legally deposited free copy it knows of, plus an oa_status naming the flavour. Gold means the journal itself is open access, hybrid means an open article inside a subscription journal, green means somebody deposited a copy in a repository, bronze means free to read on the publisher's site with no licence attached, and closed means what it says. The email is not optional; leave it off and you get an HTTP 422 saying so.

That is the whole free route. What follows is what those indexes actually contain, which is the part nobody explains and the part that decides whether a failed search means anything at all.

What each index actually covers

Every count below was taken from the service's own API, homepage or published title list on 11 August 2026. They move, so read them as scale rather than as constants.

One number out of that table is worth pulling forward, because it changes how you should read an empty result. Of the 323,998,640 works OpenAlex held on 11 August 2026, 219,369,239 carried a DOI. That leaves roughly 105 million works, about a third of the index, with no registered identifier at all. Plenty of real scholarship has never had a DOI and never will. So "this reference has no DOI" is not a finding, and "the DOI check came back empty" is not a verdict. It is a reason to search by title.

What a library adds that the free route does not

If you have institutional access, use it, and not only for the paywalls. Your tuition already pays for subject databases curated in a way no free index is, and the reason to start there is the limiter. Almost every library database carries a checkbox labelled "peer reviewed," "scholarly journals," or "refereed." Tick it before you search and the result list is peer-reviewed by construction, which saves you running any of the checks below on any of the results.

Two things it does not do, and both matter. The limiter judges the journal, not the item. An editorial, a letter, a book review or a conference abstract published in a refereed journal passes the filter, because the filter is a property of the venue. Glance at the article type before you cite. And the limiter has no opinion on whether the article exists, which sounds absurd until you are checking a reference somebody else handed you.

The other thing worth finding is the link resolver, usually a "Find it" or "Full text" button, which takes a reference from anywhere and checks whether your institution has access. Access follows your enrolment rather than your location, so logging in through your institution from abroad works fine. Different fields still live in different databases: JSTOR for long archival runs in the humanities, PsycINFO for psychology and the behavioural sciences, PubMed for medicine and the life sciences. If you are deciding where to search and how many sources you need, our guide on finding sources for a research paper walks through the whole workflow.

Google Scholar, and the one thing it will not tell you

Google Scholar (scholar.google.com) is the best free starting point for almost any topic. Three habits turn it into a real research tool.

• Search exact phrases in quotes. Quotation marks force an exact match, so "urban heat island" finds papers about that construct rather than every paper that mentions cities.
• Follow "Cited by". The count under each result links to every later paper that referenced it, so one strong older paper walks you forward into the work built on it, including anything that contradicted it.
• Connect your library. In Settings, under Library links, add your university. Scholar then shows a full-text link beside results your institution can reach.

The caveat is the one thing to internalise: Scholar indexes broadly, which is its strength and its weakness. Preprints, conference papers, theses, working papers and the occasional predatory journal sit in the same result list as refereed articles, unlabelled. Scholar finds scholarly-looking work without vouching that any of it was reviewed.

One more habit, and it is one we got wrong ourselves. Do not sort by citation count. It is the most tempting control on the page, because a heavily cited paper feels like a safe citation. The problem is that sorting by citations ranks everything in the corpus that matched your words rather than the handful of results you were shown, so the winner is whatever famous paper happens to contain one of your terms. We measured this on OpenAlex across twenty real queries on 11 August 2026 and an off-topic paper took first place on all twenty: a query about class-size reduction in schools was headed by a heavily cited paper from an unrelated field that shared one common word. The mechanism is not specific to one index, so treat the button with suspicion wherever you find it. Rank by relevance, take the first page, and judge influence yourself from the counts printed beside each result.

How to confirm a journal is genuinely peer-reviewed

Finding a scholarly-looking article is not confirming it was refereed. These checks get more definitive as you go, and the free ones come first on purpose.

Look up the ISSN rather than the name

Names are copyable, ISSNs are registered. Put the ISSN into https://portal.issn.org/resource/ISSN/ and compare the website the record names against the site you are actually on. If they differ, stop. That is the check that catches a cloned journal and it takes about thirty seconds.

Ask DOAJ, which will answer in one line

For open access, DOAJ admits journals only against published criteria, so presence there means somebody looked. You can ask it about a specific journal without going through the site: https://doaj.org/api/search/journals/issn:2041-1723 returns JSON whose total is 1 if the journal is listed and 0 if it is not. No key needed, two requests a second.

What that listing is actually worth is unusually easy to establish, because DOAJ publishes both its bar and its failure rate. Its application guide requires, among much else, at least five research articles a year, at least two independent reviewers per article, an editorial board named with institutional affiliations, and every author charge disclosed before submission, with a cap of 25% on how much of a journal's output may come from its own editors and institution. In 2025 it received 8,887 applications and rejected 64% of them. It also removes journals it has already admitted, and publishes a public log of those removals. Counting that log ourselves on 11 August 2026 gives 2,156 withdrawals recorded since February 2024, the most common reason on 1,168 of them being failure to adhere to best practice. That is our own count rather than a figure DOAJ publishes, so read the shape rather than the decimal: a directory that ejects a thousand journals for behaviour is doing work that one which only ever grows is not.

See who registered the DOI

A DOI prefix, the 10.NNNN half, identifies the organisation that registered it. There are twelve DOI registration agencies, of which CrossRef accounts for more than 90% of usage. You can ask which one owns any prefix with a single URL: https://doi.org/doiRA/10.1038 returns [{"DOI":"10.1038","RA":"Crossref"}], and it accepts bare prefixes, full DOIs and comma-separated lists. A prefix nobody has registered comes back with a plain DOI does not exist. Two caveats before you over-read this, both from CrossRef's own documentation: a prefix tells you who originally registered the DOI, not necessarily who publishes the journal now, because prefixes never change when a title transfers, and one publisher can hold several, so back-issues and current issues of the same journal legitimately carry different ones.

Read the journal's own statement

Every genuine peer-reviewed journal describes its process. Open the "about", "aims and scope" or "for authors" page and look for a plain account of how submissions are reviewed, single-blind, double-blind or open. A real journal is proud of this. A vague page, or none, is a quiet warning.

Ulrichsweb, if you can reach it

Ulrichsweb is the periodicals directory libraries use, covering 383,000 serials across 977 subject areas, and it marks refereed titles with a small referee-jersey icon. Two things to know. It is a subscription product, reachable through most university libraries and through none of the open web. And its own documentation says the icon is only shown once its staff have confirmed the review process, so a missing icon means unconfirmed rather than not refereed. Read an absence there as a question, not an answer.

That ranking is not a flourish. It is how our own system grades the same three signals internally, and it exists because a resolving DOI kept being mistaken for a verified source. To run the identifier half of it on something specific, our free DOI lookup shows the record a DOI is actually registered to, which is the fastest way to see that an identifier belongs to a different paper than the one you were handed. No account needed.

What peer review is actually worth

Time for the opinion this page is really built on, because it changes how you should use everything above. Peer review is a filter, not a guarantee, and treating the phrase "peer-reviewed" as a synonym for "true" is the mistake that most often survives a careful search.

Two measurements make the case better than any argument. On timing: an analysis for Nature of every paper in one large index carrying both submission and acceptance timestamps found a median of around 100 days from submission to acceptance, a figure that had barely moved in thirty years. That is the number to hold against a journal claiming a two-week turnaround. On accuracy: a randomised study of 607 reviewers, who were sent manuscripts with nine major errors deliberately planted in each, found that reviewers spotted a mean of 2.58 of the nine, and training barely moved it. Fewer than a third of the errors somebody had put there on purpose.

So what is peer review for? It establishes that people with relevant expertise, and no stake in flattering the author, looked at the work and did not object. That is a real and useful floor, and it is why your assignment asks for it. It is not a certificate. It means the claim has been checked, not that the claim is correct, and the practical consequence is that "peer-reviewed" earns a source a place in your argument rather than immunity inside it. You still read the methods. You still check the sample. You still notice when a 2011 finding has been contradicted twice since. A student who treats the label as a floor writes better papers than one who treats it as a ceiling, and markers can tell the difference from the sentences alone.

When you hit the paywall

Your library first, because that is what the subscriptions are for. After that the open routes, in an order that matters more than most guides admit.

Unpaywall gives you a list of locations and nominates one as best_oa_location. Ignore the nomination. The "best" copy is frequently the publisher's own, which is precisely the one most likely to refuse the request, and occasionally it is a citation stub rather than the paper. Read the list yourself and try repository copies first: PMC mirrors open most reliably, then university and institutional repositories, then the publisher. CORE and Europe PMC are worth a look when Unpaywall comes back empty, because they aggregate deposits Unpaywall has not linked.

Two URL tricks that save more time than they should. On arXiv, the abs/ page is a wrapper around the paper rather than the paper: swap abs for pdf in the address and you have the full text. On bioRxiv and medRxiv, append .full.pdf to the article URL. And when a publisher page refuses you outright, look at the address it refused you on, because publisher landing pages carry the DOI verbatim in the path, in the shape .../doi/full/10.1080/…. Pull the DOI out and hand it to Unpaywall. That is worth doing rather than giving up: in one capture of ninety-four direct page fetches from our own pipeline, sixteen came back refused, and almost every one of those had a free copy a single Unpaywall hop away.

How often is a free copy there at all? The STM Association's open access dashboard puts gold open access at 40% of articles, reviews and conference papers published in 2024, up from 14% in 2014, with subscription-only down to 54% from 70% over the same decade. Add green copies deposited in repositories and your practical odds of opening a recent paper for nothing are close to even. Older papers are a different matter, and that gap is what a library subscription is really buying you.

Predatory journals and lookalikes

Not every journal that calls itself peer-reviewed is. A 2019 consensus of scholars and publishers from ten countries, writing in Nature, settled on a working definition: outlets that prioritise self-interest at the expense of scholarship, marked by false information, a lack of transparency, and aggressive solicitation.

Be careful with the numbers, because the honest position is that nobody knows the scale. The figure everyone quotes is Shen and Björk's estimate of around 420,000 articles in predatory journals in 2014, up from 53,000 in 2010. It is twelve years old and it has no successor: no one has repeated the measurement.

The cheapest tell is on the article itself, and it is why the 100-day median above is useful. Many of these journals print "received" and "accepted" dates. Received on the 3rd, accepted on the 9th, published on the 14th is not review, it is invoicing. The other thing to watch for is a hijacked journal, a fraudulent site cloning a real title's name, ISSN and branding, which is the exact failure the ISSN lookup at the top of this section catches and the reason it is at the top. The three frauds behind the phrase "fake journal" are quite different from each other and need different checks, and we take them apart, with the lists worth using and the ones that are frozen, in how to tell if a journal article is fake.

Peer-reviewed is not the same as real

Here is the trap that catches careful students. Every check above confirms that a journal does real peer review. None of them helps if the article you are holding was never written, and that is exactly the gap AI walks into. Ask a chatbot for peer-reviewed articles on your topic and it will produce them: real-sounding journals you can confirm are refereed, plausible authors, the right kind of title, a correctly formatted DOI. You can look the journal up, find it genuinely refereed, and still be holding a reference that points at nothing, because the journal is real and the specific paper is invented.

The share of fabricated references is not small. A 2023 study in Scientific Reports found that 55% of GPT-3.5 citations and 18% of GPT-4 citations were entirely fabricated, for reasons baked into how these models work. So add the check that confirming peer review never covers: make sure the article exists. Paste the DOI after https://doi.org and confirm it resolves to a paper whose title and authors match what you were given. Search the exact title in Google Scholar or your library. An article whose journal is genuinely refereed but which appears nowhere except the citation that named it is the most dangerous kind, because it sails past every peer-review check on the journal's reputation alone. The full method is in how to check if a citation is real.

For a whole list rather than one reference, paste it into our free citation checker, which looks every entry up in CrossRef and OpenAlex and reports two things separately: whether the identifier resolves, and whether the record it resolves to describes the work you cited. Fifty entries at a time, no account. Once you have a source you want to keep, the citation generator turns its DOI into a finished reference in APA, MLA, Chicago, Harvard, IEEE, Vancouver or numbered, with every field taken from the publisher's registered record rather than retyped. What none of them answers is whether the paper supports your sentence, which no tool that has not read the paper can, and which we cover in when a source says something else.

Peer-reviewed is also one kind of credible source rather than the whole of it: a dataset or a working paper by a leading researcher can carry a claim, and a refereed paper can still be wrong for your question. How to find credible sources weighs authority, currency and purpose, and how to read a research paper covers the skill that tells you whether the article in front of you supports your sentence.

Peer-reviewed and still nonexistent

This is the part of the problem CiteOwl is built around. It is verify-first: it searches the same academic indexes described above, the same OpenAlex and CrossRef and DOI infrastructure you would use by hand, reads the papers it finds, and cites only sources it actually retrieved, with the supporting quote attached to each one. A peer-reviewed-looking but nonexistent article never makes it in, because there was never a real paper to read. You still decide whether a source fits your argument. What goes away is the one failure no peer-review check catches, the reference that looks impeccable and is not there.