Policy

Artificial Intelligence Use Policy

Artificial intelligence assists in producing this site, and we do not hide it. This page explains where AI fits in our process, what human review covers, and where responsibility lies.

Where we use AI

We draw on artificial intelligence tools in researching content, drafting and editing text, and developing the site's infrastructure. We say so openly, because we believe readers have a right to know how what they read was made. For us, AI is a production tool: a powerful successor to the printing press, the spell-checker and the search engine. The power of the tool does not reduce the need for oversight; it increases it.

Human review and editorial responsibility

AI-assisted production is not exempt from our editorial process: before publication, content is checked against the standards of our Editorial Policy and Source and Evidence Policy. Responsibility for everything published rests entirely with us; "the AI wrote it" is not a valid excuse on this site. Our commitment to readers concerns the outcome, not the method: every piece of information here is held to the same standard of accuracy, whoever or whatever drafted it.

Known risks of AI, and our safeguards

We know that AI models can produce information that does not exist, mix up sources, and be wrong in a confident voice. Our safeguards against these risks are: factual statements are checked against reliable sources; where a figure is uncertain, we use "approximately"; our ban on invented studies and fabricated institutional attributions applies fully to AI-assisted content; and in sensitive areas such as health, the bar for review is higher still. Even so, we cannot reduce the chance of error to zero, and we say that honestly.

If you find a mistake

If you spot an error in our content, do not worry about whether it came from a human or a machine; either way it is our error, and it is fixed through the same process. Reporting and correction are described in our Corrections Policy.

A working version

This policy is a working version, and since AI tools and good publishing practice around them are changing quickly, it is likely to be updated more often than most. Any significant change in our approach will be noted on this page.

First published: 2026-08-12Last reviewed: 2026-08-12Editorial status: working editionReport an error