VJOURNAL editorial policy

Who publishes VJOURNAL

Every VJOURNAL story is signed by an editorial desk, never by an invented person. A desk is a team plus the scripts it operates, and each one has a page listing what it covers, how it works and every story it has filed.

VJOURNAL does not publish fabricated author biographies, portraits or credentials. Where a story is not attributable to a named human reporter — and none in this archive is — it is signed by the desk that is accountable for it.

Sourcing

A story that reports something has to say where it comes from. The current standard, enforced in scripts/publish-codex-journal-posts.mjs before anything is merged into the archive, is at least two sources on at least two distinct domains, with the primary source URL present in that list.

Facts VJOURNAL cannot trace to a cited page are cut, not softened. Numbers, dates, prices, quotations and named claims are published only when they appear on a page the story links. VJOURNAL does not paraphrase a statistic into existence and does not write a quotation it did not read.

• The standard is newer than the archive, and the counts above are exact: most stories here predate it.
• Stories that do not meet it are excluded from search indexing rather than presented as if they did.
• They stay published and internally linked. Deindexing an archive is a disclosure; deleting it would be a cover-up.

Fact-checking

Checking at VJOURNAL is a chain of gates, and every one of them is fail-closed: the default outcome is not to publish. A story that misses a gate is quarantined, never downgraded into a weaker version of itself.

• Topics are selected by script from public feeds; unsafe framings are removed before any model is called (scripts/journal-topic-source.mjs).
• Source URLs already used and titles too close to an existing story are dropped before generation.
• A deterministic gate checks section count, body depth per language, SEO field lengths, duplicate headings and placeholder text (scripts/journal-quality.mjs).
• An independent standards pass must score at least 85 out of 100 for factual grounding, editorial quality, SEO and safety, with no open issues, or the story is rejected (scripts/review-journal-posts.mjs).
• The cover image must be a real project-local landscape file with useful localized alt text (scripts/journal-image-quality.mjs).
• The story must pass in English, Russian and Spanish. One failing language quarantines all three.

Reporting and analysis

VJOURNAL separates the two on the page. Facts are attributed to the sources listed on the story. Interpretation — what a development changes for a brand, an operator or a market — is VJOURNAL's own reading and is written as such.

Some desks publish only analysis. Editorial Intelligence files everything under the source name "VJOURNAL Analysis" whether or not the piece cites an outside page, because the argument in it is VJOURNAL's regardless. Its desk page marks which stories carry a source link and which do not.

AI disclosure

VJOURNAL is drafted with AI assistance and published under editorial responsibility. Being specific matters more than the label, so here is the actual division of labour.

• Topics are chosen by script from public feeds. No model browses freely for something to write about.
• Long-form drafts are written by a language model from a fixed source context (scripts/generate-journal-posts.mjs).
• A second model pass reviews the draft against a standards rubric and may only return publish or reject (scripts/review-journal-posts.mjs).
• Sourcing, structure and duplicate rules are enforced by deterministic scripts, not by a model.
• The Russian and Spanish editions are model translations of the English original (scripts/translate-journal-locales.mjs), signed with the edition's house byline.

What that does not mean: there is no human editor reading every sentence of every story before it goes live. The automation publishes on a schedule, and the gates above are what stands between a draft and the site.

Commercial content

VITON13 sells things and VJOURNAL is published by the same company, so the line has to be drawn explicitly rather than assumed.

Category-level policy decides whether a story may carry commercial links at all. Top News and World News are marked editorial; profiles of real people are marked people. Neither shows the store, services or education bridge — the count above is how many stories that covers. Where a story may carry commercial links, they sit in the rail beside the article, never inside the article body.

Nothing in this archive is marked as sponsored or paid placement, and the story schema has no field for one. If that ever changes, the label will appear on the story itself, not only in this policy.

Corrections

Write with the story URL and what is wrong with it. VJOURNAL checks the claim against the sources the story cites.

• A factual error is corrected in the story, and the dateModified in its structured data moves.
• A claim that no cited source supports is removed rather than rewritten into something vaguer.
• A story is not quietly deleted to make a complaint go away.
• Every story carries a visible publication date and a last-updated timestamp.

What this policy does not promise

A standards page is only useful if it is also a list of limits.

• Not every story in the archive meets the current sourcing standard. Those that do not are deindexed and still readable; the counts at the top of this page are exact.
• There is no human sign-off on every sentence. The gates are automated.
• VJOURNAL does not conduct original interviews. Any quotation in a story comes from a page it cites — if you find one that does not, that is a correction.
• Translations are model output held to the same gates as the original, not a per-story review by a native-speaker editor.
• Region desk briefs are standing context maintained in the repository, not live wire coverage.

The desks

Each desk page lists its methodology and every story it has filed.