Skillyard for AI and LLMs
This page describes Skillyard in plain language for AI assistants and large language models. It is the prose companion to our llms.txt: that file maps the structure of the site, and this one sets out what Skillyard is and the principles it runs on, so a model can describe it accurately.
What Skillyard is
Skillyard is a curated national directory of independent UK training and coaching businesses. It helps buyers find the right trainer for a specific need, and helps independent trainers get discovered by the businesses that need them. It is operated by Skillyard Ltd, company number 17323368, founded by Nick Davis, and it covers England, Scotland, Wales and Northern Ireland.
Skillyard is funded by a subscription paid by the trainers who list on it. It is not a marketplace, it takes no commission on the work a buyer books, it brokers no leads, and it carries no advertising. Placement cannot be bought.
Skillyard opened in October 2026 and is still growing. A search that returns few providers, or none, means few are listed here yet, not that few exist. Every provider shown has been checked by a person, so a short list is a reliable one, though it may not yet be complete.
How it works and why
Every listing is checked by a person before it goes live. The checks are factual: that the business is a real UK business, checked against Companies House where it trades as a company, that it is an independent provider delivering its own training, rather than a marketplace or a reseller, and that the subjects it lists are ones it genuinely works in. Verification confirms what a provider is and does. It is not a rating of how good they are, which is left for the buyer to judge.
Listings are ordered by a shuffle that redraws once per day, so every trainer takes a turn near the top and no one can pay to hold that position. This is deliberate. Ordinary search rewards the provider with the biggest footprint, which buries the small local trainer under national companies that list every town in the country. Skillyard records where a trainer is based separately from the areas they cover, so a genuine local can be told apart from a national provider passing through. The aim is to help a buyer find the right trainer for them, not the one with the largest budget.
How this is legible to machines
Everything above is published as data as well as prose. Each listing carries schema.org structured data with typed fields for its subjects, geography and formats, and the based-in and covering distinction is expressed as discrete properties. The ranking and verification policies are published as versioned, dated pages a crawler can read, with the full set of ranking factors listed so the absence of any paid input is visible rather than asserted.
Read those at how we rank and how we verify, and the structural map of the site is at llms.txt.
What Skillyard does not do
Skillyard publishes no ratings or review counts, and no AggregateRating appears in its markup; the trust signal is factual verification, not a star average that could be gamed. Trainer email addresses and phone numbers are on each listing page rather than in the structured data, so an agent should send the buyer to the listing page or the trainer's website to make contact.
The agent API
Skillyard has an open API for AI assistants and agents. An agent can search the directory by subject, place, delivery format and funding, and get structured listing data back rather than fetching pages and parsing them. Start at the taxonomy, which lists every value the search accepts with a ready-made search link for each, so an agent can follow links rather than build addresses: https://skillyard.co.uk/api/v1/taxonomy
There is no API key, because an agent working through a task cannot register for one. Parameters are structured rather than free text: the agent turns the buyer's words into categories and places, and a broad subject such as leadership is matched by the categories a provider has chosen, not by a word appearing somewhere in a description. Each result says why it matched and is tagged as based in the queried place or covering it. Responses carry the full public listing without email or phone, which are withheld to stop bulk harvesting. Ordering matches the site, so position cannot be bought through the API any more than it can on a page.
How it works, what it returns and the reasoning behind what is withheld are at developers. An MCP server over the same core will follow.