Authorship is restricted to people. An AI system cannot be listed as an author, cannot hold responsibility for a claim, and cannot be corresponded with. Every listed author must have contributed to the work and must be able to defend it.

The submitting author warrants that all listed authors have approved the submission and that no contributor who meets authorship criteria has been omitted.

Disclosure is mandatory and specific. State which systems were used, which versions, at which stage — hypothesis generation, data collection, analysis, drafting, reference assembly — and what the human authors verified independently.

Undisclosed AI involvement in the substance of the work is grounds for rejection at any stage and for retraction after publication. A general statement that 'AI tools were used' is not disclosure.

Submissions are screened for textual overlap with published work and with the author's own prior output. Substantial undisclosed overlap is grounds for rejection.

Fabricated data, fabricated references, and image manipulation are treated identically regardless of whether a human or a system produced them. Responsibility rests with the authors.

Every reference with an available DOI must carry it. Identifiers are resolved against the public registry, compared field by field with the citation as printed, and checked for retraction or withdrawal.

Corrections are returned to the authors rather than made silently, because the reference list is the authors' statement about their evidence.

Every article carries a data and code availability statement naming a real, reachable location, or explaining specifically why the material cannot be shared. 'Available on request' is accepted only where a legal or ethical restriction is named.

Where AI systems produced results, prompts, model versions, parameters, and the number of repetitions form part of the material to be made available.

Work involving people requires a named ethics approval or an explanation of why approval was not required, plus a statement on consent and anonymisation.

Do not submit identifiable personal data. Where personal data was processed by an AI system, say which system and under what terms.

All financial and non-financial competing interests must be declared, including employment by or equity in organisations whose systems are studied in the work.

Funding sources are stated with grant numbers where they exist.

Published articles are not silently edited. Errors are addressed by a dated correction notice attached permanently to the article page, an expression of concern where an investigation is ongoing, or a retraction notice where the findings do not stand.

Retracted articles remain visible with a retraction notice across the page. The record of what was published is not deleted.

Any editorial decision can be appealed once, in writing, free of charge, on the ground that a report misread the manuscript. Appeals are assessed against the submission thread.

Complaints about the conduct of the journal itself are answered in writing, and the response is published where the complaint concerns a published article.

Each accepted article receives an immutable platform identifier that is never reissued or reassigned, and a permanent page.

We are honest about the limits of a small platform's permanence guarantee: we operate no external preservation agreement, so authors should retain their own copies of their accepted manuscripts.

All published content is readable in full, at no cost, without an account. It is not available for download, and printing is suppressed.

This is a licensing position and a deterrent, not technical protection: anything a browser can display can be captured. Reuse beyond reading requires written permission from the authors.