Our crawler
User agent: IslandGraphBot/1.0 (+https://island.gg/crawler)
Island Graph is a directory of Guernsey’s businesses and organisations, built from what each of them publishes about itself. To build it we read public web pages with a crawler. This page says what the crawler reads, how it behaves, and how to stop it.
Who is behind it
The crawler is operated by Island Graph Limited, a Guernsey company (registration no. CMP78596) registered with the Office of the Data Protection Authority as a data controller under number DPA11733. It is run by Niki Wiles. Questions go to hello@island.gg. How we handle personal data is set out on the privacy page.
What it reads
- the public websites of Guernsey businesses and organisations — their home, about, services, team and contact pages;
- the public pages of registers and directories, such as the Guernsey Registry and island business directories.
It reads public pages only. It does not log in, submit forms, or read anything behind a password or a paywall. It is scoped to Guernsey: it does not roam the wider web.
How it behaves
- It identifies itself on every request with the user agent above, so it appears in your server logs under its own name with a link back to this page.
- It reads your
robots.txtbefore fetching and honours it for theIslandGraphBottoken, falling back to the*rules, and it respects aCrawl-delaydirective of up to 60 seconds. This has been in force since 4 September 2026. - It pauses between requests to the same site, and it reads a site in full once, returning from time to time to see whether anything has changed.
- A separate, low-volume checker fetches a cited page once to confirm that a quotation shown on Island Graph really appears there. It sends the same
IslandGraphBotidentity and readsrobots.txtfirst; a page we may not read is simply not checked, and the fact it would have supported is not published.
What we do with what it reads
We publish facts an organisation states about itself — what it does, where it is, who leads it, how to reach it — and we link every such fact to the page it came from, so a reader can check it. We do not take facts from pages a site has asked crawlers not to read.
For a small business, the phone number, email or address the business gives out is often the owner’s own, so the contact details we show are the ones the business itself publishes. We make every reasonable effort not to publish a home address unless the business invites customers to it. If we have got one wrong, email hello@island.gg and we will take it down.
How to block it
Add this to the robots.txt at the root of your site and the crawler will stop at its next visit:
User-agent: IslandGraphBot Disallow: /
You can also disallow particular paths rather than the whole site. If you would rather not edit robots.txt, email hello@island.gg and we will exclude your site by hand.
How sure we are of each fact
Every fact we show carries one of four states, worked out from the evidence behind it and never from a model’s confidence score. A fact is statedwhen a page of the business’s own says it in so many words and we can show you the quote; an attributewhen it comes from the business’s Google Business Profile on a given day; inferred when we read it off a page that does not state it outright (these show muted, with a small mark and the page they came from); and withheld when we have no page to point to, or when the kind of fact must be stated to be shown at all: a qualification, a membership, an award, a founding year, a person, a parent company, a number of anything, opening hours.
The machine-readable record (the structured data on each page, and /api/entities/<id>) carries stated and attribute facts only by default, each with an evidence_status field. Add ?include=inferred to receive the inferred facts too, marked as such. Withheld facts are not served.
Corrections and removal
If something on Island Graph about your business is wrong or out of date, or you would like your listing removed, see the page for owners or write to hello@island.gg. We aim to act on corrections within two working days.