Where the data comes from, how often it changes, and what it cannot tell you.
The reference currently runs on a seed list of 798 categories, written from public knowledge of Google's taxonomy. The names and gcids have not yet been checked against Google's API, and country availability is not verified. Services shown are common editorial suggestions, not Google's predefined list. Everything is labelled as seed data until the first import from Google runs.
Editorial guidance: 798 categories written, 0 reviewed.
Category names, gcids and predefined services come from the Google Business Profile Business Information API, the same source the Business Profile interface uses. We request the full list for each country (United Kingdom, United States, Ireland, Canada, Australia, New Zealand) in English, including each category's predefined service types.
A gcid (for example gcid:plumber) is Google's stable identifier for a category. Display names can change and differ by country; the gcid usually does not.
We do not scrape third-party category sites. This site is independent and not affiliated with or endorsed by Google.
The import runs weekly. Each run is compared with the previous one, and every difference is recorded: categories added or removed, names changed, and services added or removed. Nothing is deleted. A category that disappears is marked as no longer listed and keeps its history.
The first import for each country sets a baseline and records no changes, so the change log only shows real movements after that. See the change log and the RSS feeds.
Definitions, primary and secondary guidance, common mistakes, related categories and FAQs are written by us, based on Google's published guidelines and practical local SEO experience. They are opinions about good practice, not Google policy.
Every category starts unreviewed. Unreviewed pages are visible but kept out of search engines until a person has checked them. We avoid statistics and ranking claims we cannot support.
Photos come from Pexels, are credited on every page, and are processed to one consistent look. Where no photo fits a category well, we show a plain typographic card rather than a misleading image.
Search, the Wizard, the Audit, the Stack Builder and the Bulk Mapper all run in your browser. What you type is not sent to a server. Meaning-based matching uses a small open model (all-MiniLM-L6-v2) that downloads once and runs locally. Wizard and Audit rules are written by hand and readable in the source code.