CiteSiteFree tool
AI entity analyzer
Reads a page and separates two things people usually see as one: what your markup declares about who you are and what you sell, and what a machine is left to work out for itself.
Your homepage is usually the right page — it is where an entity is declared, and the page an assistant reaches first.
The failure this is looking for
Not a missing field. Three fields that disagree.
A site declares Acme in its schema.org Organization, Acme Software in og:site_name, and Acme — project tracking for agencies in its title. Every one of those is correct, written by a different person at a different time, and none of them is wrong. To an assistant deciding whether a mention on a review site refers to this company, they are three strings that do not match.
That is the common case, and it is invisible from inside the company, because everyone there already knows the answer.
Declared and inferred, shown separately
The report is deliberately split. Names from schema.org and Open Graph are marked as declared; names taken out of your title or your H1 are marked as read from copy. So are the topics — those come from your own headings and your most repeated phrases, and they are included precisely to show how thin an inference is next to a statement.
Nothing in the output was produced by a model. There is no summarising step and no identification step, which is a constraint of the free tools and, here, the right design: a tool that guessed your entity would be demonstrating the problem rather than reporting it.
What to do about it
- Pick one name. The legal name or the trading name, but one, and use it in every declared field.
- Declare it in a schema.org
Organizationon the homepage, and reference that as the publisher elsewhere. - Add
sameAs. It is a five-minute change and it is the only part of this that recruits sources you do not control. - Name what you sell as a
Product,SoftwareApplicationorServiceon the page that sells it.
The schema checker will tell you whether the markup you add actually parses — which is a separate and unfortunately common failure.
What it does not tell you
Whether the assistants have in fact merged you with somebody else. That is a question about their answers rather than about your page, and the only way to answer it is to ask them. A full analysis does that, on a schedule, and records whether each answer came from live search or from memory.
Questions
- What is an entity, in this context?
- The thing behind the name. When an assistant reads “Acme” on your site, in a review, and in a forum thread, it has to decide whether those are one company or three. An entity is the answer it settles on, and everything it later says about you is said about that answer rather than about your website.
- Why does it not use AI to identify the company?
- Because that would answer a different question. A model given your homepage will usually work out who you are — that is what the paid analysis does. This tool checks whether a machine has to. What it can read unaided is the whole point: if the only route to your name is inference, then inference is what every assistant is doing, and inference is where two similar companies merge.
- My name is in the title and the H1. Is that not enough?
- It is what most sites rely on, and it usually works. It is also the weakest form of the claim: a title is a sentence written for people, it changes when marketing changes it, and it is read positionally rather than as an assertion. A schema.org Organization states the same thing in a field whose meaning does not depend on where it sits.
- What is sameAs and why does it matter so much?
- A list of other URLs that are the same entity — your LinkedIn, Crunchbase, GitHub, Wikipedia entry. Each is a thread tying a mention of your name somewhere else back to you. Without them, a declaration is a name in a vacuum; with them, an assistant has corroboration from sources it already trusts.
- Which page should I check?
- The homepage, first. That is where an entity is normally declared and the page an assistant reaches first. Then whichever page you most want associated with a specific product, because a Product or SoftwareApplication node lives on the page that sells it.
Related
Schema markup for AI search: the types that actually get read — why Organization is the one type that earns its place on almost every page.
Declaring who you are is the half you control. CiteSite measures the other half — whether ChatGPT, Gemini and Perplexity name the right company when someone asks.