GEO Audit Tool
Can ChatGPT, Perplexity, and Google AI find and cite your website?
Paste your URL and get your AI Visibility Score in 30 seconds.
Where AI engines get their information
From three layers, not one. A technical audit reads one and a half of them, and it is worth knowing which.
Crawled data
Your indexed pages and what everyone else writes about you. They decide which category a model files you under, and who it compares you against.
Feeds and APIs
What you publish as structured data: prices, availability, attributes. It decides how you are compared when someone asks for a recommendation. No free audit sees this, because it lives in your feed rather than on your page.
Live site data
Real price, stock, reviews, checkout. They decide whether an agent can act on a buyer's behalf, or merely describe you.
This audit tells you whether you can be cited, not whether you are
Those are two different questions with two different answers. A site can pass every check on this page and never appear in a ChatGPT answer, because access is a precondition and not a cause. The only way to know whether you are cited is to ask the engines, repeatedly, and count the answers that name you.
In the citability model we run inside Anomalia, this technical audit carries 10% of the weight. The rest is extractability (25%), evidence density (25%), entity clarity (20%) and third-party corroboration (20%).
Frequently asked questions
What is GEO, and how is it different from SEO?
SEO optimises for a position in a list of links; GEO optimises for being named inside a generated answer. The unit of measurement changes: not average position, but the share of answers your brand appears in. A site can rank first on Google and never be cited by ChatGPT.
Does this audit prove that ChatGPT cites me?
No. It checks that AI crawlers can read you, that structured data is present, and that the page is intelligible to a machine — the preconditions. Whether you are actually cited is a different measurement, one that requires querying the engines repeatedly and counting.
Where do AI engines get their information?
From three layers: crawled pages, the feeds and structured APIs you publish, and live site data. The three can contradict each other — a price in the feed that differs from the page — and contradiction is one of the fastest ways to stop being recommended.
Does my ecommerce site need Product and Offer data?
Yes, and it is the most expensive gap this audit finds. Without a Product block containing an Offer that states price, currency and availability, an assistant can describe the brand and cannot rank a single item. Add an identifier (sku, gtin) and the product URL and it can also send you a buyer.
Is llms.txt actually worth doing?
It is an emerging standard, not a requirement: no engine has declared it a ranking factor. It costs ten minutes and hands engines a curated map of the pages that matter, so we recommend it — but it carries little weight, and if you could fix only one thing it would not be this.
What actually moves citation?
The one controlled study in this space (Princeton, KDD 2024) measured the visibility lift of individual tactics in generative engines: direct quotations roughly +41%, statistics +33%, fluency +29%, citing sources +27%, while keyword stuffing did nothing at all. Engines have moved since; the ordering — evidence density beats keyword tactics — has held in every replication.