How LeraLink Uses AILeraLink Team

How LeraLink Uses AI

A fixed set of tasks, proposals until a person decides, and policies per installation

Last updated: 13 November 2026

What AI does

AI at LeraLink has a fixed set of tasks: translating; suggesting search-engine fields for a work; making a feature image; building the search index; naming and describing the concepts and people that published texts are about; and, where it is enabled for an installation, a composition assistant that makes proposals only.

What AI does not do

AI does not write, title, summarise, research or evaluate a work for its author. It does not change a published text, approve a translation, decide on rights or give anyone a role.

Proposals until a person decides

A translation, a suggestion or an image is a proposal, marked as made by a model, until a person decides on it. Two exceptions. Machine translations can be shown to readers before a person has reviewed them; they are then labelled "AI translation · Not yet human-reviewed". And suggested search-engine fields that nobody has reviewed go live with the work, marked as unreviewed, and the Studio keeps asking for a review. Concept descriptions are computed from published text, are marked as machine-made, and a field a person has fixed is never written over by an automatic process. Once a person accepts a generated image, readers still see a small sign that it was made with AI, and the file itself carries machine-readable metadata saying so.

Which models

Each task tries its models in a fixed order set by the platform; an installation's administrators can see that order, with each provider, on the page "AI in this installation". The providers are Cloudflare Workers AI, OpenAI and Google. The platform chooses the models; an installation writes editorial instructions, such as how to translate, but never a prompt or a model.

AI policies

Each installation is under one policy:

  • Standard: the platform's models in their usual order.
  • EU only: only models hosted in the EU and recorded as not training on what they receive. The installation's text is left out of the search index; keyword search keeps working.
  • Open models only: only models whose weights are published.
  • No external AI: only models running on Cloudflare, where LeraLink itself runs.

A model about which a required fact has not been recorded does not count as meeting a rule.

What an installation can see

Its administrators see, under "AI in this installation", each task, whether it is on, and the models that would run, in order, with each provider's country, hosting region, whether it trains on what it receives, and how long it keeps it.

Data sent to providers, and their terms: Privacy Statement, section 4.