KoiPaper
RegisterLog in
English
EnglishEspañolالعربية简体中文

HOW EVIDENCE WORKS

A recommendation should show its reasoning.

See what informs the shortlist, how conflicting requirements are handled and where the evidence stops. AI explanations support a fixed ranking process.

Start with your manuscript

How a shortlist is built

  1. 01

    Extract the manuscript fields

    Identify the question, methods, subject and language from the available text. Missing information remains visible.

  2. 02

    Retrieve directory candidates

    Use the current DOAJ directory and available fields to compare subject, language, fees and submission conditions.

  3. 03

    Look for related publications

    Query Crossref, OpenAlex and Europe PMC with source fallbacks. A failed query is reported as unavailable evidence.

  4. 04

    Apply condition checks and scoring

    Code applies known constraints and a fixed weighted ranking. AI explains the result; it cannot override a condition check.

  5. 05

    Present the available evidence

    Return up to five qualifying journals, with recommendation order, reasons, available sources, unknowns and next checks. Fewer candidates are shown when necessary.

Different sources answer different questions

01

Directory facts

DOAJ provides journal identity and directory-reported fields. A directory entry does not establish that every current submission requirement is met.

02

Publisher evidence

Official journal pages help confirm scope, fees and author instructions. A topic profile inferred from metadata is not an official scope statement.

03

Public publication metadata

Crossref, OpenAlex and Europe PMC support related-record searches. Metadata and journal identity matches have limitations; a missing record is not proof that no article exists.

04

Your input & AI explanation

Your requirements guide the search. AI provides explanations of available fields, rather than a new source of journal facts.

A missing fact never becomes a confirmed match

Pass

Available information supports the condition.

Conflict

A known requirement is incompatible: for example, a confirmed $900 fee exceeds a $500 budget.

Unknown

A required fact, such as the current fee, has not been confirmed. It is shown for follow-up.

A ranking method you can interpret

Recommendation order combines subject fit, recent related records, manuscript language, article type, budget and evidence completeness. Known hard conflicts are excluded. Missing information weakens support; it is not treated as a confirmed match. These initial experimental rules still require calibration. The order is not an acceptance probability or validated accuracy claim.

What related publications can—and cannot—tell you

Recent related articles show a journal’s publication history, not successful submission samples or a promise of acceptance. Topic relevance uses text matching and has limits. Low-relevance results are not presented as supporting records. Publisher policies can change; confirm the current instructions before submitting.

Questions worth asking

What if fewer than five journals qualify?

The report shows the actual number rather than adding weak candidates to fill five places.

What happens if a database fails?

The other checks can continue. The report shows the unavailable source or missing evidence instead of presenting the query as successful.

What should I verify before submitting?

Open the official journal links and confirm the current scope, manuscript type, language, fees and author instructions. Unknown conditions still need verification.

Explore data coverage ↗