How We Use AI

We cover artificial intelligence closely, so it would be strange not to say plainly how we use it ourselves. This page exists because a reader deserves to know how what they are reading was made.

The short version

A person is accountable for everything published here. AI is a tool in the production of our articles, in the same category as a search engine or a spreadsheet: useful, fast, and incapable of being responsible for anything. Responsibility sits with the people named inour bylines, and with the newsroom when a piece is published under its name.

Where we use it

• Drafting. Assembling structure, first drafts and explanatory passages from material we have gathered.
• Research assistance. Summarising long documents, filings and reports so an editor can find what matters faster.
• Editing. Tightening prose, catching repetition and flagging unclear passages.
• Production. Headline and summary variants, tagging suggestions, and internal tooling that has nothing to do with what you read.

Where we do not

• We do not publish invented facts, figures, quotes or sources.A language model will produce a plausible statistic on request, and a plausible statistic that nobody measured is a lie with a decimal point. Numbers in our articles must be traceable to a real source.
• We do not fabricate reporting. We will not claim to have interviewed, surveyed or spoken to anyone unless we did. Analysis is labelled as analysis; it is not dressed up as fieldwork.
• We do not invent people. No fictional bylines, no composite reporters, no borrowed identities. Every name on this site belongs to someone real. See oureditorial policy.
• We do not let a model decide what is true. Verification is human work, and it is described in ourfact-checking policy.

What we record internally

Our CMS records how each article was produced — written by a person, AI-assisted and human-reviewed, or drafted by AI and not yet reviewed. We keep that record from the moment a piece is created, because reconstructing it afterwards is not realistic, and because we would rather be able to answer the question than guess at it.

We are not yet displaying that label on individual articles. When we do, it will be because the record is complete and accurate enough to be worth showing, not as a badge.

When we get it wrong

An error is an error regardless of how it was produced. We do not treat "the model wrote it" as an explanation, because it is not one — a person chose to publish it. Errors are corrected and disclosed under ourcorrections policy, with no separate, softer treatment for mistakes that originated in a tool.

If you find something on this site that looks fabricated — a figure without a source, research that cannot be located, a quote you doubt — tell us atinfo@geniuscreative.us and we will check it. We have removed our own work on those grounds before, and we would rather do it again than leave it up.

This will change

These tools are moving faster than any policy written about them. We expect to revise this page, and the date at the top is the last time we did. If our practice changes, this page changes with it.