Everything we publish about getting named by AI assistants is on this page — the method, the rubric, the files and the measured data. Almost none of it is gated, because a document behind a form is invisible to the engines it is about, and we would rather be quoted than counted.
Every document exists as a web page first, then as a PDF to read and Markdown to copy from — the same content three ways, never a fuller version behind the form. The facts below are stated here as text on purpose: a page that links PDFs without repeating their substance in HTML is the exact pattern our own checker fails a customer for.
| Resource | Formats | What it costs you | Updated |
|---|---|---|---|
| The GEO Field Guide | Page · PDF | Nothing | When engines change |
| The page checklist | Page · PDF | Nothing | When a check changes |
| The sample client report | Page · PDF | Nothing | When the report template changes |
| The templates pack | Page · PDF | Nothing | When a crawler token moves |
| The GEO glossary | Page · PDF | Nothing | As the vocabulary settles |
| The Visibility Index dataset | Page · CSV · JSON | Nothing; an address for new editions | Each measured edition |
| The five-part course | Five emails | An address | Evergreen — it ends after lesson five |
| The scoping builder | Interactive · Markdown | Nothing; a form only for a reply | As the method changes |
Everything in this group is published in full on the page and downloads as a file without asking you for anything. The "email me a copy" forms are a convenience for people who want it in their inbox, never a gate.
The complete method: how assistants decide who to name, the six layers of visibility in dependency order, how to decide what to write, how to build a page that gets quoted, and how to measure without fooling yourself. About 45 minutes end to end.
Page · PDF · optional email copy
All 26 checks our free checker runs on a page, in the five weighted groups that make up the score, each with what it means and the fix. This is the entire rubric — including the part about what a page score cannot tell you.
Page · PDF · optional email copy
The client-ready AI-visibility report, section by section, filled with clearly-invented example data — and annotated with the client question each section answers. Built for the most common job in this discipline: reporting what the assistants say to the person who asked.
Page · PDF · optional email copy
robots.txt rules per crawler, an llms.txt skeleton, the Organization, Product, Article and FAQ structured-data blocks, a page skeleton and a fact-sheet template. The crawler tokens come from the same registry our access checker grades against.
Page · PDF · optional email copy
Every term you will meet in AI-visibility work, defined precisely — from parametric memory to owned-citation rate — because sloppy vocabulary is how bad advice spreads.
Page · PDF
How every number we publish is produced: which engines we call and how, why one answer is an anecdote, how the judge is calibrated against human labels, and the arithmetic of the Visibility Index in full. It ends on the limitations, because a methodology with no limitations section is marketing.
Page · PDF · no account, no form
Eight questions about your site, your stack and your publishing capacity, and out comes a prioritised plan in dependency order — including the work your situation lets you skip. Computed in your browser from fixed rules, so nothing in it can be invented.
Interactive · Markdown download · form only if you want a reply
Two resources where an address earns you something an anonymous download cannot: a sequence that unfolds, and a dataset that gets a new edition.
Five short emails on how assistants decide who to name and what actually moves it. No jargon, one idea each, and it ends after lesson five — that is a promise, not a funnel.
Five emails · one-click unsubscribe in every one
Which brands assistants actually name, by category — measured on a fixed suite of non-branded buyer questions, with the 95% confidence interval, sample size, engine list and run id travelling with every row. Published as CSV and JSON for you to re-analyse and argue with. A category with no completed run is simply absent rather than estimated.
CSV and JSON · address only if you want the next edition
These two are built around your own information, and asking for one is a genuine signal that you are evaluating something. That is the only place a full form is a fair exchange — and the underlying findings stay free either way.
Your own free check, written up for the person who is going to ask "so what?" — the score in context, what each finding costs, what to do first, and a plain statement of what a page score cannot tell you. Run a check, then follow "Get the written report".
Name, company and email · save as PDF · the findings themselves are free
The scoping builder gives you the plan on screen. Leave your details with it and someone who has actually read your answers replies — about your stack, your capacity and your market, rather than a template.
Name, company and email · your answers travel with the request
Score any URL against the 26 checks, with the findings ranked and a shareable result link.
Whether each AI agent can actually reach your site — robots.txt parsed per agent, plus live probes that catch the CDN and firewall blocks robots.txt will not show you.
Type a search, see whether Google shows an AI Overview for it, which sources it cites, and whether your site is one of them — from a live Google result, not an estimate.
Valid Organization, Product, Article and FAQ JSON-LD from a short form. Nothing invented — it only emits what you type.
A tidy index of your key pages, built from your sitemap. Cheap insurance, not a lever.
Because we would be contradicting ourselves. Our own published method says a gated PDF is invisible to crawlers, that gating baseline material defeats the visibility work it describes, and that it produces weak leads into the bargain. A company selling AI visibility that hides its methodology behind a lead form has already failed its own audit.
So the rule we apply to ourselves is the one we give customers: gate only where the exchange is fair and the request is itself a signal. Knowing what to do is free. Knowing whether an assistant now names you — and whether the change you shipped moved that number — takes repeated sampling across engines over time, and that is what we sell.