Methodology

Every number on this site is reproducible from open data. This page describes how, including the judgement calls and the things we decided not to publish.

1. Which journals are included

We start from every source in OpenAlex that is a journal, and require it to:

• have an ISSN;
• have more than 200 indexed articles, which excludes stubs and one-off titles;
• have published since 2022, so the list reflects active venues.

That leaves about 50,500 journals across all disciplines.

2. How we decide what counts as biomedical

OpenAlex assigns each journal a set of topics, and each topic rolls up to a field and a domain. We weight every topic by how many of the journal's articles it covers, then measure what share sits in the Health Sciences domain or in the biomedical parts of Life Sciences (biochemistry and molecular biology, immunology and microbiology, neuroscience). A journal is included at a share of 0.45 or above.

We weight rather than taking the single top topic because top-topic assignment is noisy — The Lancet's leading topic is filed under Engineering. That leaves11,447 journals.

The threshold is a trade-off we tested. Lowering it to catch a handful of general medical journals pulls in roughly 1,800 food-science, aquaculture and plant-biology titles, which is a much worse outcome. Instead a short, explicit allowlist of ten flagship journals that OpenAlex mis-files — The Lancet among them — is added by ISSN. That list is in the source and is deliberately small.

3. Excluding data artefacts

OpenAlex contains a few aggregate records where unattributable works are collected. One claims 5.5 million articles; another claims 2.8 million articles and 14 million citations under the name of a small entomology journal. These are not journals, and left in they would top every ranking on the site. We drop any record claiming more than 600,000 articles.

4. Merging in DOAJ

For journals OpenAlex flags as DOAJ-listed, we look up the DOAJ record by ISSN and take the fields nothing else provides: article processing charges with currency, fee waiver availability, declared peer review model, licence, plagiarism screening, preservation, and average time to publication. That covers 3,620 journals.

5. Metrics, and what they are not

The metrics shown — h-index, i10-index, and 2-year mean citedness — are OpenAlex citation statistics.

They are not Journal Impact Factors. The JIF is a licensed product of Clarivate, computed from a different citation database over a different set of source items. We do not publish JIF values and the numbers here should never be quoted as one. Where an institution requires an impact factor, you need Journal Citation Reports, not this site.

6. Verification signals

Each journal page carries a panel of checks. Every one is a fact drawn from a public record — is it in DOAJ, does it declare a review model, does it screen for plagiarism, does it archive, does it assign DOIs, is it still publishing.

There is no score, no grade, and no verdict. A journal missing several checks is not thereby labelled anything; the reader is given the facts and draws the conclusion. Many legitimate subscription journals sit outside the open-access registries entirely and will show few confirmed checks.

7. What we deliberately do not show

Founding years. OpenAlex's first-publication field is distorted by mis-dated references: it puts PLoS ONE at 1806, Cell at 1814 and Nature Communications at 1800. Some values are correct — the New England Journal of Medicine really did begin in 1812 — but there is no way to tell which, so we omit the field rather than print numbers we cannot stand behind. We show the most recent indexed article instead, which is reliable.

Acceptance rates. Very few journals publish them and those that do define them inconsistently. A figure that cannot be compared is worse than no figure.

Machine-written journal descriptions. We could generate a paragraph of prose for every journal. We do not, because it would introduce unverifiable claims into pages whose entire purpose is verifiability.

8. Similar journals

The suggestions on each journal page are computed from shared-topic overlap: each journal becomes a weighted vector over its topics, normalised to unit length, and we take the nearest by cosine similarity. It reflects what a journal publishes, not its prestige or its publisher.

9. The matcher

The journal matcher projects your abstract onto the same topic space and ranks journals by overlap. Topic terms are weighted by inverse document frequency so that specific words count for more than common ones, and the whole computation runs in your browser — your text is never transmitted.

Overlap says nothing about whether a journal would accept your paper. It is a shortlisting aid.

10. Updates

The dataset is rebuilt from source rather than edited in place, so figures move when the upstream records move. Anything you read here is a snapshot, and the journal's own site is always the authority on its current fees.

Sources and their licences are listed under data sources.