Transparency
AI Disclosure
Last updated 26 August 2026
The short version
AI tools are used here for drafting, editing, and code. They are not used to decide what is true. Every factual claim published on this site was checked by a person against a source, and that person is accountable for it being right.
Where AI is used
- Drafting and restructuring prose, which is then edited.
- Copy-editing: grammar, tightening, headline options.
- Writing and reviewing the code this site runs on.
- Summarising long primary documents — earnings transcripts, regulatory filings — as a reading aid. The claims that end up in a post are then read back against the original document.
Where AI is not used
- Generating statistics, benchmark results, or any number presented as fact.
- Producing quotes. Every quote is taken from a document or recording that is linked.
- Publishing anything unreviewed. No post goes live without being read end to end by a person.
- Manufacturing volume. This site is deliberately small, and posts are removed when they stop earning their place.
Why this page exists
A language model will produce a confident, plausible, wrong number if asked for one, and will attribute it to a real organisation. That failure mode is the reason measured figures on this site are labelled as measured and cited figures are attributed to a named source — see the methodology page. It is also why a card that computed a fake “impact score” from headline keywords and displayed it as a percentage was removed from every post in August 2026.
If something here reads like it was generated and not checked, it is a bug. Report it via the corrections policy.