The short version
We use AI as a drafting and research assistant. We do not use it as the final author of any published piece. Every article on FinNotes is reviewed by a named human editor before publication, and the named author and editor are accountable for the content regardless of which tools contributed to the draft.
Where AI sits in our workflow
We treat AI the way we treat a junior analyst: useful for first drafts, data lookups, summarisation of long documents, and structuring rough notes into prose. The output of that work is always handed to a human writer or editor before anything reaches a published page.
Specifically, AI tools may be used to:
- Generate a first draft from an editor's outline and source material.
- Summarise long official releases, central-bank statements, or earnings transcripts.
- Suggest headline variants, deks, or rephrasings of unclear sentences.
- Translate primary-source quotes from non-English documents (always cross-checked against the original).
What AI does not do
- Final publication. No AI output is published without substantive human edit and review.
- Fact-checking. Every statistic, quote, and source citation is verified by a human against the primary source. AI is not trusted to validate facts.
- Inventing facts. We do not generate fictional people, quotes, sources, or numbers. If a piece references a person, a price, or a document, that reference is real and verifiable.
- Editorial judgement. Decisions about what to cover, how to frame it, and what conclusions to draw belong to the named author and reviewing editor.
Why we don't label every AI-touched paragraph
Some publishers tag each AI-assisted sentence. We don't, and we think the per-sentence label gives a false signal of precision — AI tools touch most modern editorial workflows at some stage (autocomplete, grammar suggestions, summarisation), and tagging the visible output paragraph implies the rest came from elsewhere, which isn't true either.
Our standard is at the article level instead: the named author and editor are responsible for the published article, regardless of which tools contributed to the draft. If a piece misstates a fact, the author and editor are accountable; "the AI wrote it" is not an excuse we accept internally and not one we would offer publicly.
Reviewed-by signature on every article
Every published article surfaces both the writing author and the
reviewing editor in the byline — visible to readers, and present
in the article's schema.org/NewsArticle structured
data via the author and editor fields.
That double signature is the explicit promise that a human
reviewed the piece before publication.
Authors and editors maintain individual conflict-of-interest disclosures, reviewed regularly. See our conflicts policy for the personal-investment restrictions that apply across the masthead.
Corrections and updates
If we publish an error of fact — whether the source was AI-assisted
or not — we correct it. Corrections are made inline with a clear
"Correction" prefix, and the article's dateModified
is bumped so search engines and AI citation layers see the change.
See the full corrections policy for details.
Questions
If you spot a factual error, or have a question about our use of
AI in a specific piece, write to
[email protected] with the article URL and the
specific claim you want us to check.