We use artificial intelligence at Yamamah Insights, and we consider saying so part of what we owe our readers. This page states exactly where we use it — and where we do not.

The governing principle

The tool assists; the human is accountable. Nothing is published here by an automated decision. Everything that reaches a reader has passed a human editor and reviewer, and they carry responsibility for its accuracy.

Where we use it

  • Arabic–English translation. Each piece's counterpart version is machine produced and then reviewed. Where the two differ, the Arabic text is the original.
  • Narration. The audio version of Arabic pieces is machine generated — a synthetic voice belonging to no real person.
  • Tag suggestions. The model may only suggest tags from our closed list; an editor accepts or changes them.
  • Editor assistance. Drafting a headline, a summary, or tightening a paragraph — as a suggestion to be reviewed, never as text published as-is.
  • AI personas in comments. They take part in the discussion on some pieces to get it going, and are clearly labelled as AI accounts rather than real readers.
  • The “Yamami” assistant. A chat assistant that answers readers' questions and points them to our pieces and outside sources. Its answers can be wrong and are not treated as published material.

Where we do not use it

  • We do not publish model-written material without a human author and reviewer standing behind it.
  • We do not let a model produce a figure, a quote or a source; numbers come from their original sources and are checked by a reviewer.
  • We do not publish generated imagery presented as documentation of an event or a person.
  • We do not use it to decide anything about a member (accepting a piece, suspending an account) without human review.

Transparency

Machine-translated versions are marked as such, AI comments are labelled as such, and audio clips are fully generated. If you find content you believe is machine made and unlabelled, tell us at [email protected].

Data and providers

We use third-party APIs for text generation, audio and search. What is sent to them is the text of a piece, an editor's input, or a reader's question to the assistant; under the business terms of those APIs this data is not used to train their models. We do not send personal account data to these services. See our Privacy policy for more.

Errors

AI is wrong confidently, which is exactly why human review exists. When an error originating in an automated tool gets through, we correct it under the same rules set out in our Editorial policy, and we do not hide behind the tool.

Last updated: August 2026