How an entity audit works
An entity audit runs in six steps. Agree which entities matter, inventory every name and identifier in use, review the schema.org graph on your key templates, check profiles and knowledge bases such as Wikidata, test how AI assistants and search engines describe you, then turn the findings into a prioritised fix list.
1. Scope: which entities matter
On the first call we agree what is in scope: usually the organization itself, each location, the founders or key people, and the main products or services. Each gets a single intended identity: the canonical name, the canonical web page that defines it, and the identifier its structured data should use.
2. Entity inventory
Every name, spelling, legal name, former name, branch label and handle in use is collected from your website, structured data, social and business profiles, directories and knowledge bases, and from the answers AI assistants give. The result is one table showing which variants are legitimate alternates and which are mistakes that split the entity. Background: how name variants split a business.
3. Schema graph review
The JSON-LD on your home page, about page, location pages, author pages and key content templates is parsed and checked:
- Is there one Organization (or LocalBusiness) node with a stable
@id, referenced everywhere, or several competing ones? - Do WebSite, WebPage, Article, Person and Service nodes point to it correctly, with no references to undefined
@ids? - Do the properties match the visible page and Google's documented requirements, and is there review markup that isn't eligible?
- Do multiple sites you run describe the same entity with the same identifiers?
The type-by-type rules used are documented on Schema Institute, and you can run the same browser-based checks yourself with the schema.technology validator.
4. Profiles and knowledge bases
The profiles that identify you elsewhere are checked for name, address, phone, website and description consistency, and against what your sameAs lists. For Wikidata, the audit records whether an item exists, whether it is about you, and whether you meet the notability policy with serious, publicly available references. If you don't, the report says so rather than recommending an item likely to be deleted (why that matters).
5. AI assistant and search recognition
A fixed set of questions about the business, its people and its offerings is put to major AI assistants and search engines. The answers are recorded with the sources they cite, so each wrong name, outdated fact or confusion with another entity can be traced back to the page or profile it came from.
6. Fix list
Findings are ranked by how much confusion they cause and how hard they are to fix, and each one comes with the specific change: the corrected JSON-LD, the profile field to update, the page to edit. See deliverables.
What you need to provide
Access to the website (or its HTML), a list of the names, locations and sites you know about, the profiles you control, and anything you already know is wrong. No passwords to third-party profiles are needed for the audit itself.