DOI Download: How to Find a Paper’s Free Full Text Legally

Figure 1, explained — every tool, site and organisation in the diagram

Figure 1 is the whole method on one page, and this section takes it apart: what each step does, what every tool is for, and a link to every service, repository and organisation it names, so you can go to the source rather than take our word for it. The steps it draws are the four checks in this guide, opened out into six.

The six steps in the figure

Every tool named in the figure

Thirteen services appear on the figure’s tool cards, plus the resolver in step 1, two more named in the research pathways, and the organisations behind them. Each entry below gives the figure’s own description and the address to go to.

The organisations behind the figure

• FRELIP — the publisher of this guide, and the site the figure’s footer sends readers to: frelip.org.
• NNFA — the Nigerian & African Knowledge Access Initiative, FRELIP’s partner initiative, credited in the figure’s header and footer and the co-author of this guide.
• The DOI Foundation — the body that governs the identifier system used in step 1: doi.org.
• Creative Commons — behind every cc-by and cc-by-nc label in the figure’s licence table: creativecommons.org/licenses.
• Sci-Hub — named in the figure only to be refused. See Why we do not link to Sci-Hub.

The tables and lists in the lower half

The lower half of the figure is working lists rather than a diagram, and each one maps onto a section of this guide:

• Research pathways — five columns (journal article, book or chapter, thesis or dissertation, biomedical research, African research), each a route of five or six sources. This is the figure’s version of the four checks, adapted to what you are looking for.
• The six common answers — open at publisher, repository copy, accepted manuscript, preprint, metadata only, nothing found. The same six verdicts set out in check 4 below.
• Access vs version vs licence — three separate questions: can I read it, what am I reading, what may I do with it. Expanded under version and licence below.
• Free to read ≠ free to reuse, and nothing found ≠ nothing exists — the two boundaries the figure is careful to draw, both treated below.
• FAQ, researcher’s checklist, key sources — the four questions researchers ask most, a six-point checklist, and the list of links the figure was built from.

Below is the box-by-box detail for the elements that need more than a line: the resolver, the indexes, and the repositories and institutions that held copies in the live queries behind this guide.

Box by box: the resolver and the indexes

The institutions named in the worked examples

The first example, 10.1126/science.aac4716, is the 2015 Science paper on the reproducibility of psychological science. It is paywalled at the publisher and open in repositories — and its location list is the best teaching case in this guide, because the copy the index calls “best” has no PDF at all. These are the institutions our live queries returned, each linked to the repository or university that holds the deposit:

Two numbers in the figure are worth pausing on. Unpaywall returned 15 locations for the Science paper; OpenAlex returned 47, of which 10 carry a PDF. That gap is not a contradiction — it is the same database presented at different depths, which is the subject of check 3.

The four checks, in order

1. Resolve the identifier. Confirm the DOI is real before you chase anything.
2. Ask the index. Unpaywall, or OpenAlex for the full location list — the same data either way.
3. Ask a genuinely different source. CORE, then the place the work actually lives: Scholar, the biomedical archives, or the preprint server named by the prefix.
4. Read the answer. Six possible verdicts, each with a different meaning for what you may cite and reuse.

Everything below uses one DOI so you can follow along: 10.1126/science.aac4716, paywalled at the publisher and green in repositories — the most common situation you will meet.

Check 1 — Resolve the identifier

A DOI has two parts: a prefix that begins 10. and a suffix chosen by the registering organisation. Paste the whole thing after https://doi.org/ and the system looks up the current address and forwards you there. We tested it: https://doi.org/10.1038/nature12373 answered with a redirect straight to the Nature article page.

If nothing loads, the identifier is wrong, truncated, or mistyped — and this is where researchers lose time. FRELIP learned the lesson while building its own DOI resolver: a malformed DOI that fell through to a title search came back with a different paper, complete with a confident verdict and a plausible title. A silent wrong answer is worse than an error, because you cite it.

So: fix the DOI, do not substitute a search. A DOI that fails to resolve is a typo, not a paywall.

Check 2 — Ask the index, and know what it is not

Unpaywall is the fastest direct answer to “is this free, and where?”. Paste this into your browser, replacing the DOI and your own email:

https://api.unpaywall.org/v2/10.1126/science.aac4716?email=you@example.org

The email is not optional: Unpaywall requires one as a condition of use, and a request without it returns HTTP 422 with the message “Email address required in API call”. We confirmed both responses. If you would rather not touch an API at all, the free Unpaywall browser extension puts the same answer on the article page you are stuck on.

The part most guides get wrong. Unpaywall and OpenAlex are not two independent databases. They are run by the same team, and Unpaywall now runs on OpenAlex data — the same index delivered through two interfaces. OpenAlex simply shows you more of it: 47 locations where Unpaywall lists 15. So the honest way to use them is:

• Use either one to answer “is there an open copy?”. You do not gain confidence by asking both.
• Use OpenAlex when you need the full location list, versions and citation counts.
• Treat CORE and the subject repositories as the second opinion. That is where an independent answer actually comes from.

And read the fields properly:

• is_oa — true or false.
• oa_status — gold, hybrid, bronze, green or closed.
• oa_locations[] — walk the whole list, not just best_oa_location. This is the single most useful line in this guide.

Check 3 — Ask a genuinely different source

Our hero example proves why. Unpaywall labels the “best” location Tilburg University, version publishedVersion — and url_for_pdf is null. Meanwhile the copies that actually contain a PDF are the acceptedVersion and submittedVersion deposits at Groningen, the LSE, Dundee, Keele and Edinburgh. A researcher who stops at best_oa_location concludes there is no file. A researcher who reads the list has six.

Then widen to sources that are not the same database:

• CORE — https://api.core.ac.uk/v3/search/works/?q=doi:"10.1108/10650750710776422". That identifier is closed on Unpaywall and on OpenAlex, and CORE returns a PDF.
• Google Scholar — the [PDF] link beside a result is usually an author-posted copy, and Scholar indexes places the publishers’ indexes do not.
• PubMed Central and Europe PMC — for funded biomedical work, deposit is often a condition of the grant.
• The preprint server named by the prefix — 10.48550/arXiv.*, 10.1101/* for bioRxiv and medRxiv, 10.31235/* for SSRN-style deposits. If the DOI itself is a preprint registration, go straight to the server.

One practical warning about CORE. Its open interface works without a key at low volume, and it throttles: a single lookup answered normally in our tests, while repeated rapid calls returned HTTP 429. A 429 is not “no copy exists”. It means the service declined to answer, and you should try again later rather than conclude the paper is closed. If you plan to look up many DOIs, register for a free key instead of hammering the open tier.

Check 4 — Read the answer: six verdicts

Once you have asked, the answer is one of six. They are not variations on yes and no: each tells you something different about what you are reading and what you may do with it.

Version and licence: what you may actually do with the file

Version literacy is the part that most “install Unpaywall” posts flatten. Three habits cover it:

• Cite the version you opened. Pagination, figures, copy-editing and sometimes wording differ between the accepted manuscript and the published article. Quoting a sentence an editor later reworded will not match the copy your examiner opens.
• Read the licence, not just the verdict. A copy under cc-by can be reused and redistributed with attribution. A copy marked other-oa, or with no licence at all, is typically free to read, not free to redistribute. Our hero example’s “best” copy is other-oa; two of its PDFs are cc-by.
• Bronze is not permanent. A bronze copy is free to read on the publisher’s site with no open licence, which means the publisher can take it down. If the file matters to your work, keep it.

For the colour vocabulary itself — gold, green, diamond, hybrid — see Open Access Explained. For how to cite what you found, see How to Cite Sources.

When the answer is “no free copy”

A negative is only as strong as the sources you asked, and coverage is uneven. The clearest example we found is in the figure: 10.48550/arXiv.1706.03762 is a real DOI for a famous preprint, and both Unpaywall and OpenAlex returned HTTP 404 for it — not “closed”, but not found at all, because that registration route sits outside their index. The paper is free to read at arXiv.

Before accepting a negative:

• Preprint servers — arXiv for physics, mathematics and computing; AfricArXiv for African scholarship; bioRxiv and medRxiv for life and medical sciences; SSRN for the social sciences.
• The author’s institutional repository — most Nigerian universities now run one. FRELIP has written about how they work: Institutional Repositories in Nigerian Universities.
• The author’s own page or profile.
• Interlibrary loan or your library’s document delivery. If you are at a Nigerian university, this is often the real route to a genuinely closed paper — slower than the internet, entirely legal, and underused.

And then the oldest method: email the author. Two lines are enough — who you are, what you are working on, which paper you would like to read. It works more often than you think. One caveat worth stating plainly: an author may share the accepted manuscript freely, but many publisher agreements restrict posting or sending the typeset version, so treat the reply as whatever the author is entitled to send, not as a blank licence.

Why we do not link to Sci-Hub

Sci-Hub reproduces and distributes copyrighted articles without permission, which is unlawful in Nigeria and in most other jurisdictions, and it has been blocked or litigated against in several. FRELIP does not link to it and does not recommend it.

The Nigerian position is not vague. The Copyright Act 2022 (Act No. 8 of 2022), which came into force on 17 March 2023, keeps exclusive reproduction and making-available rights with the rights holder, and its fair dealing provision — section 20(1)(c) — covers non-commercial research and private study, judged against purpose, the nature of the work, how much was used, and the effect on the market. Reading is not the problem. Operating a library that replaces the market for the work is. The full text of the Act is on WIPO Lex.

Our full method list, including the routes that need no tools at all, is in How to Download Research Papers for Free.

Questions researchers ask

How do I download a research paper for free using its DOI?
Paste the DOI after https://doi.org/ to confirm it resolves, then ask Unpaywall (or OpenAlex — same data), then ask CORE or Scholar or the preprint server named by the prefix. If one reports an open copy, download it and note the version and licence.

Are Unpaywall and OpenAlex independent checks?
No. They are the same team and now the same data: Unpaywall runs on OpenAlex. Asking both is one census asked twice. CORE, the subject repositories, Scholar and the preprint servers are where an independent answer comes from.

Unpaywall says there is an open copy, but there is no PDF. Why?
Because best_oa_location ranks by preference, not by whether a file exists. Our worked example’s “best” copy is a landing page with url_for_pdf: null. Walk oa_locations[] and you will find the actual files.

Why does my DOI open a page but not the full text?
Because a DOI identifies a record, not a readable file. It takes you to wherever the publisher keeps that record; whether the full text opens depends on whether that particular copy is open access.

Is it legal to download a paper from a DOI?
Yes, when the copy is open access or the publisher permits it, which is the case for everything described here. Downloading from a pirate service is not legal, and it is unnecessary for the large share of paywalled papers that do have a legal free version somewhere.

What happened to the Open Access Button?
It was retired. The service closed on 18 November 2025 and now points users to Unpaywall. Some older guides still recommend it, including one of ours, which we are correcting.

CORE returned HTTP 429. Does that mean there is no copy?
No. It means CORE declined to answer because the open tier is throttled. Retry later, or check Scholar and the subject repositories. Never record a throttled lookup as a negative finding.

What is the difference between a DOI and a URL?
A URL is where something is right now and it breaks when a site is rebuilt. A DOI is a permanent identifier with registered metadata, and its registrant is obliged to keep the link current. Our explainer on what a DOI is covers the difference in full.

How FRELIP helps

The four checks above are the manual version of what FRELIP builds for: making legal, open scholarship findable without a paywall, an account or a fee.

• How to Download Research Papers for Free — twelve legal methods, for when you have a topic rather than a DOI.
• How to Find Academic Sources — judging whether a source is worth citing in the first place.
• Open Access Explained — the colours, in plain terms.
• What is a DOI? — the other half of this topic: getting and citing one for your own work.
• Open-access journal feeds by discipline — browse current open research in your field.
• FRELIP research services — literature search support for students and lecturers.

FRELIP fact-check report

Sources