How this site is made

This page was last updated 2026-10-14 (14 October 2026). Most numbers here are computed by scripts from public data. Event records are drafted with the help of an AI model and are published only after a named editor approves them. Nobody on this site is a made-up person, and no AI model approves anything.

Computed from public data

These are produced by code at build time from the source registry, IMF PortWatch series and the approved event archive. They are not written by a person or by an AI model, and each number carries a source line generated from the source registry.

  • Charts, key facts and the measured effect of each event, using the published effect method.
  • The facts list, the day counters on ongoing disruptions, recovery time, closest calls and the home page digest.
  • The port, chokepoint, lane and country pages: their figures and lists come from the registries and the data, and the sentences around them are fixed templates that fill in those values.
  • The share images used when a page is linked elsewhere. They are drawn by code from the same data, with a fixed font and layout. They are not AI-generated pictures.

Where AI assistance is used

  • Event records. The first draft of an event record (summary, timeline, sources, severity basis) has been written by an AI model, Claude from Anthropic, working from sources it finds and cites. A script then checks each draft: required fields, dates in order, a working link and an archive snapshot for every official and press source, and no run of 12 or more words copied from a source. The draft stays in a private review folder and is not part of the site.
  • Detection. Software flags unusual moves in the daily series and possible disruptions. A flag is only a prompt for a person to look.
  • Weekly editions. A script generates the draft from the data. A person edits it and approves it before it is published or emailed.
  • Blog posts, explainers and the site itself. The blog, the learn explainers and the site code are produced with AI assistance. They are published under the name World Freight Monitor, and unlike event records and editions they do not pass through the approval gate described below.
  • News. The news pages are not written by us or by AI. Each item is a headline, a short excerpt taken from the publisher's own feed and a link to the original article. We add no commentary.

Who approves what is published

No event record and no weekly edition appears on the site without approval. The approving editor reads the record against a written checklist (editorial rules), then runs the promotion tool by hand at a terminal and types a confirmation. The tool refuses to run from a script or scheduler. The site build rejects any record without an approval stored in it, and a record approved under the test-only editor handle cannot build into the live site.

Approvals are made under an editor handle, not a legal name. At the time of writing there is one live handle, which belongs to the site's owner and operator. We do not publish invented authors, biographies or credentials, and none exists here. Pages are published under the organisation name, World Freight Monitor.

What this does not guarantee

AI models can make mistakes. The checks above catch missing sources, broken links and copied text. They cannot prove every sentence is right, and we have not measured an error rate. No lawyer, shipping professional or other expert has reviewed the records. This site is a record of public information, not advice on shipping, trading, insurance or investment, and measured effects show timing, not cause.

Corrections and reporting errors

If you find a mistake, email corrections@worldfreightmonitor.com with the page address and what is wrong. Every change to a date, number, place, cause, severity or source after publication is dated and listed in the corrections log and on the record. See also about and methodology.