Editorial policy
What this publication will and will not do, written plainly enough to be held to.
Evidence class on every article
Every article carries one of ten evidence classes, displayed above the text with its meaning spelled out. It says what kind of claim the article is reporting: a peer-reviewed result, a preprint that nobody outside the author list has checked, an agency announcement, a measurement, a model, an interpretation, a proposal, something planned, or the publication's own editorial guidance.
This is the single most important thing on the page. Astronomy reporting fails most often not by stating falsehoods but by promoting a provisional claim one level of confidence, and the class exists so that promotion has somewhere to be caught.
Sources
Primary sources, named and dated. The full hierarchy is in the sourcing policy.
Every article ends with numbered references. Each carries the publisher, the publication where there is one, the date, the URL, and the date the source was actually consulted. Where a claim rests on a paper, the reference points at the paper — not at coverage of it.
The use of software in drafting
Articles here are researched and drafted with software assistance, and this policy will not claim otherwise. Claiming human-only authorship that cannot be guaranteed would be exactly the kind of unverifiable assertion the rest of this document exists to prevent.
What that changes is nothing about the standard. Every factual claim traces to a source that was read. Every citation is real and resolves. No quote, study, author, observation or result is ever invented. A fabricated citation is easier to produce than a real one, which is precisely why the build refuses to publish an article whose references do not balance against its text.
Imagery
Scientific imagery comes from the missions and institutions that produced it — NASA, ESA, ESO, STScI, NOIRLab and Wikimedia Commons where the licence is compatible. Every image records its credit, its source page, its licence and its dimensions.
Artist's impressions are labelled as artist's impressions, prominently, because they are illustrations of an idea rather than pictures of a thing. Documentary imagery is never generated. Unrelated stock photography is never used to illustrate a scientific result.
Corrections
Corrections are published, not edited away. A substantive correction appears on the article itself, dated, describing what was wrong — and in the corrections log.
Typographical fixes are made silently. A change to a scientific claim, a number, an attribution or a conclusion is never silent.
The relationship to the encyclopedia
AsteriaStar's encyclopedia is the canonical reference: what is known about an object, with provenance on every value. This publication is the reporting: timely, authored, dated, and argued.
Articles do not duplicate encyclopedia prose. Where the two touch, the article links to the reference page instead of restating it. That keeps the reference canonical and keeps the two from competing for the same reader or the same search result.
What is not published
Claims that cannot be traced to a source that stands behind them. Results attributed to unnamed scientists. Sensationalised framing of provisional findings. Coverage assembled from other people's coverage.
None of these is a hard rule to follow. They cost stories, which is the point: a publication that never declines a story is not applying a standard.