Editorial standards
How we decide what to publish, what we use AI for, and what happens when we get something wrong.
What is in place, and what is not yet
We would rather be exact about this than sound impressive.
In place today
- A person must approve every story before publication.
- Unapproved drafts cannot reach public pages, search or the sitemap.
- Stories can list their sources and show dated corrections.
Being built
- Automated monitoring of sources and story discovery.
- A research view that keeps evidence beside each draft.
- A formal review queue and publishing calendar.
Sources
We prefer primary sources: specifications, official documentation, incident reports, papers, and statements from the people involved. When we rely on someone else's reporting, we credit it. We link to our sources so that you can check us.
Fact checking
Before a story runs, its factual claims are checked against those sources by the person approving it. When something is uncertain or disputed, we say so rather than smoothing it over.
How we use AI
AI tools help us find, organise and draft material. They make mistakes, including confident ones, so they do not get the final word. AI cannot publish on this site: a person has to approve each story, and that person is accountable for it. We do not invent quotes, sources or people.
Corrections and updates
If we get something materially wrong, we correct the story and add a dated note saying what changed. We list corrections on the corrections page. Small fixes such as typos do not get a note. If you find an error, please tell us.
Conflicts and sponsorship
If a story involves a sponsorship, a financial relationship or another conflict of interest, we disclose it on that story. Sponsors do not get to decide what we cover or what we say.
Images
We prefer original diagrams, real screenshots and official imagery, used with the proper rights. If we use a supporting photograph, it will not be presented as showing something it does not.