Do I need a Wikipedia page to show up in AI answers?

No. Most small businesses do not meet the notability criteria and do not need one. Consistency across the site, Google Business Profile, LinkedIn and official directories, plus local press mentions, is enough and within your control.

What about directories that show old data and do not respond?

Prioritise: the ones that appear in the first results for your name matter; the rest, less so. Request corrections through their forms and, in parallel, strengthen the sources you control — if 8 sources agree and one differs, systems go with the majority.

Entity architecture: one version of the truth about your

Three prices, two addresses, one lost customer

Ask an AI assistant about a business and watch what it does when it finds contradictory information. The website says "consultation 150 lei". Google Business Profile: "free consultation". A business directory: the old address, from before the move. Facebook: a different phone number. The answer turns vague ("prices vary, check directly") or, more often, recommends a business whose facts are clear.

AI systems do not rank pages; they build a representation of the entity — your business as an object with attributes: name, what it sells, where, for whom, at what price, who runs it. When the attributes contradict each other, confidence drops and the entity is skipped. Entity architecture is the discipline of controlling that representation.

What an entity is and why it matters

For Google, an entity is a unique "thing" in the Knowledge Graph: a person, a company, a place, a product. A clinic is an entity; its website is only one of the sources about it. Language models reason the same way: they accumulate statements about a name from every source they see and weigh them by how often and how consistently they appear.

The practical consequence: you do not optimise a page, you optimise the coherence of every place you appear.

The canonical description

Start with a single paragraph, written once, used everywhere:

[Exact name], [type of business], in [city, country], offers [main services or products] for [whom]. [Concrete differentiator]. [Reference price or range]. Founded in [year] by [person].

The facts that must be identical

Name — one form only. Not "Nova Dent", "Nova Clinic" and "Nova Dental Clinic SRL" in turn. The legal name can appear, but marked as such (legalName).
Address, phone, email — identical, including the format.
Hours — and holiday exceptions, updated everywhere at the same time.
Services or products — the same list, the same names.
Prices — if you publish them, one source of truth, with everything else pointing to it.
People — founder, doctors, specialists: the same name and the same role on the site, on LinkedIn and in articles.
Identifiers — the tax ID and trade register number where relevant. They tie the online entity to the legal one in public sources (in Romania: the Trade Register, ANAF, and directories such as listafirme.ro or termene.ro that republish official data).

sameAs: telling machines "this is me too"

In the Organization (or LocalBusiness) block in JSON-LD, the sameAs property is a list of URLs to your official profiles: Google Business Profile (the Maps link), LinkedIn, Facebook, Instagram, YouTube, Wikidata if an entry exists, your trade register page. It makes systems link the profiles to the same entity, instead of seeing five independent sources that might be about different companies.

The reverse matters just as much: from every profile, a link to the site — and to the same canonical URL (one form, with or without www).

Disambiguation

"Nova" is also a furniture store, a driving school and a cosmetics brand. If your name is common:

Use the full form with the specifics consistently ("Nova Dental Clinic Cluj").
Put the city and the field in the homepage title and in Organization.
Tie the entity to unique people (the founder, with their own page and LinkedIn profile).
Do not describe yourself with terms nobody else uses for you.

The facts page

A public, simple page with every fact in one place: /about or /facts. Visible text, not just JSON-LD. It contains: the canonical description, address and hours, the list of services with reference prices, the people with their roles, the legal identifiers, links to the official profiles, the date of the last update. For an AI assistant looking for "who is X and what do they do", it is the ideal page to cite. For you, it is the source you copy everywhere. And llms.txt, if you have one, should point to it in its first lines.

Wikipedia and "reference" sources

Wikipedia and Wikidata carry great weight for the Knowledge Graph and for language models. But Wikipedia has strict notability criteria and does not accept promotional articles: for most small businesses a Wikipedia page is not realistic, and forced attempts get deleted. Wikidata is more accessible for factual data (name, type, headquarters, website, founder), if the entity has independent sources. Focus on what you control: the site, Google Business Profile, LinkedIn, official directories, local press.

The 5-step plan

Audit — search for your business on Google, Maps, ChatGPT, Perplexity, Gemini and in directories. Note every fact that differs. Usually dozens come up.
Canonical description and fact list — an internal document, approved by the owner, with the single version of every fact.
Fixes at the source — site (text and JSON-LD), Google Business Profile, LinkedIn, Facebook, directories. Request corrections where you have no direct control.
Facts page and sameAs — published, linked from the footer, referenced from Organization and from llms.txt.
Monthly ritual — repeat step 1 with the same 10–20 questions; any change of fact goes through the full checklist.