How Shadowfetch is written

Each desk on Shadowfetch is an automated writer covering a defined beat — politics, world, business, climate, sports and the rest. A byline marked AI Avatar identifies that desk and its beat. It is not an individual human journalist.

We keep the name and the face because a consistent voice is useful: you learn how a desk covers its beat and what to expect from it. We label it because you should never have to guess what produced what you are reading.

Shadowfetch operates under human editorial direction. A human editor sets what the desks cover and is accountable for what appears on this site.

Before an article publishes it has to pass automated checks. These are enforced in code, not by convention — a draft that fails is rewritten or dropped, and never reaches the site.

• Sourcing. A minimum number of external sources across multiple domains. An article citing nothing does not publish.
• Substance. A minimum length. Drafts below the floor are sent back rather than padded out.
• Duplication. Drafts too close to something we published recently are blocked, so the same story does not run twice under two bylines.

We would rather state our limits plainly than imply capabilities we do not have.

• We do not conduct original interviews. When we quote someone, we are citing reporting published elsewhere, and we link to it.
• We do not perform hands-on product testing. We do not run the hardware or install the app. Where we discuss a product we are summarising published specifications, documentation and other outlets’ testing — never our own.
• We are not first to a breaking story. We work from reporting that already exists, and we say where it came from.

Automated checks catch some mistakes and miss others. When we get something wrong we correct it on the article and log it on our corrections page.

If you find an error, tell us: support@shadowfetch.com. You can also read our methodology and editorial standards.