Editorial policy · prototype
PubMeta is building a free, source-first research publication. These standards govern editorial work in the current prototype; they do not imply that a candidate has been verified, reviewed, or published.
We begin with identifiable primary sources and separate what a source reports from our interpretation. Metadata, abstracts, press material, and secondary coverage are leads—not substitutes for the underlying work. We label unavailable information and do not infer verification, peer review, replication, confidence, or source availability.
A brief should identify the study design, population or system, measured outcomes, comparison, and material limitations. Causal language requires a design that supports it. Statistical significance is not treated as practical importance, and a single result is not described as a consensus or replication.
Missing evidence, measurement limits, selection effects, model assumptions, alternative explanations, and generalizability constraints belong beside the headline finding. Priority scores organize internal work only; they are not evidence ratings.
People and process
Roles carry different responsibilities, but everyone shares the obligation to distinguish observed evidence from interpretation.
Contributors should work from the most direct source available, preserve stable citations, accurately represent methods and results, and state what they could not inspect. Drafts must include limitations and uncertainty in plain language. Contributors must not conceal relevant relationships or use authority, affiliation, or credentials as a substitute for evidence.
Reviewers assess the match between claims and cited material, check bibliographic metadata, inspect methods where available, and record specific concerns. A recommendation to advance means only that drafting may continue. Reviewers should challenge overstatement, distinguish factual corrections from editorial preferences, and avoid reviewing work where a conflict could compromise—or reasonably appear to compromise—their judgment.
Internal evidence briefs use numbered versions. Versions submitted for editorial review are retained as immutable history; a later edit begins a new draft version. Active editorial members can file internal change requests, and managing editors or administrators record an immutable resolution. This workflow is internal and does not publish, retract, or silently replace public material.
Before PubMeta activates public publication, it will define the reader-facing correction notice and intake process. We will not describe that public mechanism as available before it exists. The governing principle is already clear: material changes should be dated, explained, and linked to the affected version rather than erased.
Contributors and reviewers are expected to disclose financial, professional, academic, personal, advocacy, or competitive interests relevant to an assignment. A disclosed conflict may require reassignment or additional review. Disclosure does not prove bias; nondisclosure prevents readers and editors from evaluating it.
AI tools may assist with organization, transcription, summarization, or language, but they are not authors, sources, or reviewers. A human contributor remains responsible for every claim and citation. Material AI assistance should be disclosed in the editorial record and, when public publication exists, in an appropriate reader-facing note. Generated text must be checked against cited sources; fabricated citations or unsupported claims are unacceptable.
PubMeta’s editorial output is intended to remain free to read, without a subscription paywall. Access to an external research source may still depend on its owner; we do not label a source open or available without checking. Funding or sponsorship must not purchase coverage, favorable conclusions, or suppression of corrections.
The practical test
Can a careful reader see what the source says, what we added, what remains uncertain, and what changed?Explore the prototype →