CITE-FROM-DOI

Citation styles are less settled than they look. This tool is built against the style manuals, and checked against reference lists as actually printed by the journals that own each style. Those two do not always agree — and the CSL definitions that most reference managers use are a third answer again. Where they conflict, we follow the manual, then the journal, and we record the disagreement rather than hiding it.

Harvard (Cite Them Right): how many authors before “et al.”? Three sources, three answers. Cite Them Right sets two different thresholds — four or more authors for an in-text citation, but ten or more in the reference list. Several university library guides quote the in-text rule beside reference-list examples, which reads as “four” for both. And the widely used harvard-cite-them-right CSL file applies the in-text threshold to bibliographies too (et-al-min="4"), so most reference managers truncate a five-author reference that Cite Them Right would print in full. We follow the reference-list rule, because that is what this tool produces.

ASA: are page ranges shortened? The ASA guide's own journal example prints 82(1):224-243 in full. But American Sociological Review — ASA's flagship, published by the same association — shortens the second number in 55 of the 60 page ranges in a single 2014 article (212-39, not 212-239). The ASA CSL file shortens them too. Two against one, so we shorten, and we have noted what would change our mind: the ASA Style Guide itself, which neither of the first two sources actually quotes.

For most references the styles are unambiguous and every source agrees. The disagreements cluster in small, visible details — punctuation, author-list cutoffs, abbreviation, page ranges — and they are exactly the details a copy-editor notices. If your target journal publishes author instructions, they override everything here. Spot-check a few entries against a recent issue of that journal before submitting, and tell us if you find a disagreement: those reports are how this tool gets more accurate.

If you want the painful details… Every disagreement above was found the hard way — by rendering real references, comparing them line by line against what journals actually printed, and reading the style manuals when the two conflicted. That work is written up as it happened, including the false starts and the things we got wrong and had to reverse: the citation-proofing tracker. It is long, unpolished, and not written for a general audience — but if you have ever wondered why two tools format the same reference differently, it is the honest answer.

CITE-FROM-DOI turns identifiers into finished, correctly formatted citations. You give it DOIs — by pasting them or by uploading files — and it looks each one up in authoritative sources (CrossRef, DataCite, OpenAlex, PubMed & Google Books) and returns clean citations in the styles and formats you need.

You don't have to pull out the identifiers by hand: drop in unstructured files like PDFs, Word documents, spreadsheets, HTML pages, or reference exports (RIS / BibTeX / EndNote), and the tool scans them for valid DOIs automatically. Pick your output formats (Word, Excel, CSV, text, Markdown, HTML, BibTeX, RIS) and citation styles (APA, Chicago, MLA, a specific journal's house style from a catalog of thousands), and receive the results by email, as download links, or on screen.

• Manuscript submission
• Reference manager updates
• LaTeX bibliographies (BibTeX)
• Grant compliance
• CV management
• Literature search & review
• Author identification
• Productivity benchmarking

Manuscript submission. Gather the works you cite — paste their DOIs, or upload the manuscript PDF and let the tool pull the DOIs out. Choose the target journal's house style (search the journal-style library), export to Word or RIS, and drop the formatted references into your manuscript or the submission system.

Reference manager updates. Export your resolved citations as RIS and import them into Zotero, Mendeley, or EndNote for clean, consistent records. You can also upload an existing RIS or EndNote (.enw) file — the tool reads the DOIs back out and re-resolves them against current source data to refresh stale entries.

LaTeX bibliographies (BibTeX). BibTeX is its own thing — the bibliography format for LaTeX documents, not a reference-manager interchange file like RIS or EndNote. Export your citations as a BibTeX (.bib) file to cite them from a LaTeX manuscript, or upload an existing .bib file to pull its DOIs back out and refresh them.

Grant compliance. Turn on Grant reporting and Open access status (under Advanced) for CSV / Excel output: for each cited work you get its funder, its OpenAlex open-access status, and — the signal that matters for NIH and other federal public-access policies — whether a copy is actually posted in PubMed Central.

CV management. Format your whole publication list in one consistent style and export it to Word, Markdown, or HTML for your CV. With the Author formatting option (under Advanced) you can emphasize authors in every citation — for example, underline your own name (or bold / italicize it) — so your contributions stand out in a long list.

Literature search & review. Building a reading list from a database search? Drop in the results however they arrive — DOIs in mixed styles, a PubMed export, saved result pages, or loose PDF / Word / text files. The tool scans everything for DOIs, re-resolves each against the source databases, and returns one clean bibliography.

Author identification. Turn on Author ORCIDs and Author affiliations (under Advanced) to get each author's ORCID iD and credited institutions in CSV / Excel output. Use it when the question is whose work this is, not what it is — verifying a CV, or deduplicating a list merged from several searches.

Why a name is not an identifier, and what the blanks mean

Names do not identify people: a list can hold three J. Smiths, and one author can appear under a maiden name, a married name, or a different transliteration. An ORCID settles that; an affiliation confirms it — same name, same institution, usually the same person. ORCIDs cost you nothing, since they arrive with the record; affiliations slow the run, because CrossRef seldom carries them and the tool adds OpenAlex.

Expect gaps, and read them as coverage rather than as absence. Both fields are deposited voluntarily by the publisher, and neither was routine before ORCID became widely adopted around 2015 — so a blank on an older work usually means nobody deposited an identifier, not that the author lacks one. Affiliation is deposited far less consistently than authorship even now.

An affiliation is also a snapshot, not a current address: it is where the author was at publication. That is exactly what you want for institutional reporting, and exactly what you do not want if you are trying to find someone today. For a career list spanning a move, expect the column to change partway down — that is the data being correct, not inconsistent.

Because the two fields fail independently, they are strongest together. An ORCID with no affiliation still identifies the person; an affiliation with no ORCID narrows a common name to a plausible few. Selecting Citation count turns both on for you.

Productivity benchmarking. Turn on Citation count (under Advanced) for CSV / Excel output to get a citation count beside every work — your own output, a candidate's, or a comparison set — then sort by it or total it by year.

What it costs, why the databases disagree, and what to do about it

The count changes daily, so selecting it re-fetches every record and the run is slower than usual. The numbers are CrossRef's, not Scopus, Web of Science, or Google Scholar: use them to compare works within one list, never against a figure quoted from another database.

Each index counts a different universe, so there is no single true figure for a paper. CrossRef counts citations from works whose publishers deposit reference lists — most do, but deposit is voluntary, so some journals contribute nothing. Google Scholar reads highest: it also crawls theses, preprints, working papers, and books. Scopus and Web of Science are curated, counting only citations from titles they have selected, so they usually land in between — and differ from each other.

Expect the gap to be widest in fields that publish through conferences and working papers — engineering, economics, the humanities — and narrowest in clinical medicine, where nearly everything is a journal article with a deposited reference list.

Within one list every row is usually measured the same way, which is what makes the comparison sound. Two things can break that, and both are worth a glance before you rely on the numbers. A blank is not a zero — it means the registry holds no count for that DOI, which is routine for reports, datasets, and theses. And OpenAlex also publishes counts, so if a record's CrossRef count is missing, OpenAlex's can be filled in instead — a wider net, and not the same ruler as the rows around it. Turning on Metadata source(s) shows which registries answered for each row; if every row reads CrossRef, the comparison is clean and you can leave the column off next time.

Across databases it is never clean. If a number goes into a CV, a promotion file, or a review, name CrossRef as the source and give the date you pulled it — the figure will have moved by the time anyone checks.