How Journal Finder was built and tested

Beta · data release R-2026-10-22-01 (22 October 2026) · site version 1.8a.4 · each source's date is shown on the journal page

Journal Finder suggests ten journals whose published articles are closest to your title and abstract. For each journal it shows what official sources record: indexes, open-access model, fees, and how active the journal is. It does not rate journal quality, and it does not predict acceptance. This page explains where the information comes from, how the suggestions were tested, and what the tool cannot yet do.

At a glance

Where the journal information comes from

Every item on a journal card comes from a named source and carries the date we read it. If no source records an item, the card says "Not recorded". This never means "no".

Indexes are shown only as flags taken from each index's official list:

Other sources are the NLM Catalog, OpenAlex, Crossref, the ISSN register, publishers' own price lists, and national journal lists (Türkiye, Pakistan, Algeria and others). The full list is at the end of this page.

When sources disagree, written rules decide which value is shown. A case the rules do not cover goes to a person for review. When one source says a journal has ceased but OpenAlex or Scopus show recent publishing, the journal is not hidden: its card says the sources disagree (655 journals in this release). A journal is marked new when its first issue came out in the last three years, and announced when a publisher has announced it but no article has appeared yet.

Fees come only from the journal or its publisher, or from DOAJ and OpenAlex. A fee taken from a publisher's price list names the publisher and the date of the list, and links to the publisher's own price page; the list itself is not republished. A publisher that does not want its amounts shown can write to info@mrresearcher.com, and we will remove them. An amount from last year's list is marked “may have changed”; an older or undated amount is not shown, and a link to the fee page takes its place. The access model (fully open, hybrid, subscription) is taken from publishers' official lists where they publish one.

Cost for your country and institution

When you choose the country where the corresponding author works, and your institution if it is on our list, each journal shows what its publisher's own waiver or discount policy, or an open-access agreement, means for that choice. Each line names its source and date.

• Publisher policies: the waiver and discount rules that 20 publishers state on their own pages. Where a rule refers to Research4Life groups or World Bank income groups, those lists are used to apply it.
• Agreements: agreements between publishers and national libraries, consortia or institutions, taken from the publishers' and libraries' own pages (Saudi Arabia, the UAE, Türkiye, Qatar, Lebanon, Oman, Egypt, Bahrain, Jordan, Kuwait, Morocco and Pakistan so far). Member lists that publishers publish link a member institution to its consortium's agreement.
• What it cannot promise: many agreements have a yearly limit or conditions (article type, corresponding author, licence). Your library confirms whether your article qualifies.
• Your choice stays in your browser. The cost is worked out on your device; your country and institution are never sent to us.

These rules are dated 22 October 2026.

How the journal data were checked

• Automatic checks before every release. A validator checks every release. In its own self-test it catches all 23 errors planted on purpose.
• Comparison with hand-checked records. We compared 40 journals checked by hand with the database (794 item comparisons), on the data version before launch. Where both had a value, they agreed 90.6% of the time.
• Links. Links that no longer work are not shown. In the last check, 2,729 journal websites and 3,884 other links were hidden.
• Your reports. Each item on a card has a flag for reporting an error. A person checks every report against the original source; nothing changes automatically. Each month we publish here how many reports we received and how many led to a correction.

How the suggestions work

Your title and abstract are compared with the articles each journal has published. A journal's score comes from its closest articles and from shared concepts (subject terms and keywords). The launch version combines an updated engine (v1.4) with a standard text-matching method (BM25). The combination was chosen in development tests and then confirmed in a second registered test (below).

• Indexing does not change the order. The order reflects only how close a journal's articles are to your text.
• Journals that have ceased publication are hidden; you can show them.
• Your title and abstract are not stored. They are used for your search and then discarded. They are never written to a log.

How it was tested

Both tests are prospective. The tool knew only articles published up to 30 June 2026. The test questions were articles published after that date. For each article, we asked whether the journal that actually published it was among the 10 suggestions. Each plan was registered on OSF before the test was run, and each test set was used once.

First registered test: the engine (confirmatory)

We registered the plan on 10 October 2026 (osf.io/mcrzn). The test used 5,263 articles across 18 fields.

The registered claim was non-inferiority: the engine is at least as good as the standard method. That claim was confirmed. Superiority was not claimed. There were no deviations from the registered plan.

Second registered test: the launch version (confirmatory)

The plan is registered at osf.io/5tm2n. The test used 5,008 new articles across 18 fields that no part of the tool had seen. The registered claim was superiority: the launch version finds the publishing journal more often than standard text matching alone. A second claim compared it with the engine of the first test.

Both claims were confirmed. The launch version found the publishing journal 3.0 percentage points more often than BM25 (95% CI 2.2–3.8) and 2.9 points more often than the engine of the first test (95% CI 2.1–3.7). There were no deviations from the registered plan.

Other measures, not part of the claim: the publishing journal was the first suggestion for 10.3% of the articles and among the first five for 23.0%.

What these percentages mean. The tests count a suggestion as right only if it is the very journal that published the article. Many other suggestions can also be suitable. In a blind review of 300 suggestions from the first test (30 articles × 10), 81.7% were judged reasonable. That review was done by one reviewer and is exploratory.

Exploratory checks of the first test's engine (not claims)

What Journal Finder does not do

• It shows no impact factor, JCR quartile or CiteScore.
• It never labels a journal as fake, dead or predatory. It reports only what sources record.
• It does not predict whether your paper will be accepted.
• Indexing never changes the order of suggestions.
• Mr Researcher offers paid consulting and a paid publication support package. Neither ever affects the suggestions or the comparison.

Known limitations

Privacy and reporting errors

• We do not store your title or abstract.
• Your country, institution, language and guide setting are kept only in your own browser (local storage), so the page remembers them. They are never sent to us. The suggestions of your last search and the journals you tick for comparison, never your title or abstract, are kept in this browser tab until you close it, so that the browser's Back button keeps them.
• Server logs record only the time, page, status and response time. They contain no text and no IP address.
• Error reports need no account and do not ask for an e-mail address.
• If you use "Write to us" under the results, your e-mail address and message are sent by e-mail to Mr Researcher so that we can reply. The tool does not store them and does not add the title or abstract of your search. We keep the e-mail in our mailbox only as long as needed to answer you.

Releases

Who built it and how to cite it

Built by Mr Researcher, Muscat, Oman.

First pre-registration: Balouchi, A., & Mohtashami, M. (2026). Prospective evaluation of journal recommendation accuracy in Journal Finder on articles published after a fixed corpus cutoff date. OSF. https://osf.io/mcrzn

Second pre-registration (launch version): Balouchi, A., & Mohtashami, M. (2026). OSF. https://osf.io/5tm2n

Sources

DOAJ · OpenAlex · Crossref · NLM (MEDLINE, NLM Catalog) · PubMed Central · Scopus source list · Web of Science Master Journal List · Embase journal list · CINAHL coverage list · publishers' price lists (Springer Nature, Elsevier, Wiley, Sage, Oxford University Press, Cambridge University Press, MDPI, IEEE, Thieme, Wolters Kluwer, JMIR Publications) · national journal lists (TR Dizin, DGRSDT Algeria, ARCIF 2025, HEC Pakistan and others) · APA PsycInfo coverage list · publishers' waiver policies and agreement pages · consortium member lists published by publishers · national library pages · Research4Life eligibility list · World Bank income groups (CC BY 4.0) · hand checks by our team.

Journal-level metadata from DOAJ is used under its CC0 waiver; OpenAlex data are CC0. Index flags show only whether a journal is on a list; the lists themselves are not republished.