How we measure, exactly
Every number in a report should be explainable. This page is the explanation — what we test, how we sample, and what we deliberately do not promise.
1 · The readiness score (0–100)
AI Search Readiness is a structural score of one page: can an AI engine fetch it, parse it, and quote it? We check what the crawler actually receives (raw HTML, not the JavaScript-rendered view), structured data, answerable content blocks, llms.txt, robots/sitemap signals, and content depth. It is deterministic: the same page scores the same. SEO Foundation is the classic search hygiene baseline, scored separately — a page can be strong on Google and still invisible to AI answers.
2 · The visibility snapshot
We generate realistic, brand-agnostic buyer questions for your category (you can refine them with your product description, goal, and competitors), ask real AI engines, and detect in code whether the answers name your brand — or a competitor. Product aliases count for the brand (an answer that says a product name instead of the company name is still a mention).
3 · Sampling and ranges
AI answers vary between runs. A single ask would make the score jitter, so each question is asked multiple times per engine. A result like 3–4/5 is an honest range: the floor is questions where you were named in every sample, the ceiling is questions where you were named in any sample. A wide range means the engine is genuinely undecided about you — that is information, not noise.
4 · Real citation measurement
Separately from the knowledge snapshot, you can have grounded engines search the live web and report which URLs they actually cite for your buyer questions. From those citations we map where AI gets its answers in your space — your site, competitor pages, review sites, communities — which tells you where to invest next.
5 · A snapshot, not a ranking
AI visibility is not a stable rank. It varies by engine, time, and wording — so we measure it as a dated snapshot and say so on every report. The way to use it: fix what the report shows, re-run, and compare trends over weeks (the Monitor page automates this), not minutes.
6 · What we don't promise
We never promise that any AI engine will recommend you, and we don't fabricate visibility we didn't measure. When a check can't run, the report says it was skipped and why. The fixes we generate improve how reliably engines can understand and cite your pages — the structural lever you actually control.