Wikidata for local businesses: the hidden weapon of AI visibility

When you ask ChatGPT "tell me about the Hôtel du Roc in Castellane," the AI doesn't crawl the entire web to answer. It first relies on a knowledge graph — a dictionary of entities it knows about. And at the heart of that graph for local businesses, there's Wikidata.

Wikidata is the free knowledge base powered by Wikipedia. The major AI models (OpenAI, Google, Anthropic) read it to know who you are, where you are, and what you do. Without an item, you're an anonymous URL. With a clean, sourced item, you become a recognized entity.

We did this for 5 of our client sites (a hotel, two campsites, a spa, a gîte) in a matter of hours. Here's the method, and what you need to know before you start.

Why Wikidata, and not Google Business Profile or Knowledge Graph?

Google Business Profile and Google's Knowledge Graph are essential — but they're proprietary and closed. Other AI models (ChatGPT, Claude, Perplexity, Mistral) don't have access to them. Wikidata, by contrast, is open, structured, multilingual, and read by every AI model. One item, multiple channels.

In practice, a local business in 2026 needs all three:

• Google Business Profile — for local Google and Maps SEO.
• Google Knowledge Graph — for AI Overviews and Gemini (fed by GBP + schema).
• Wikidata — for the rest of the AI ecosystem.

What a minimal Wikidata item should contain

You don't need a Wikipedia page (different, and far more demanding). A Wikidata item for a local business comes down to 6 to 10 well-filled-out properties:

• Label (the exact name, in French and English at minimum).
• Short description (one sentence, in French and English).
• Instance of (P31) — the type of establishment (hotel = Q27686, campsite = Q281460, spa = Q300888, restaurant = Q11707…).
• Country (P17) — country (France = Q142).
• Location (P276) — the town (link to the Wikidata item for the town).
• Coordinate location (P625) — latitude / longitude.
• Official website (P856) — your URL.
• Stated in / reference URL — a verifiable source (your official website or a press article).

Optional but useful: founding date (P571), star rating (P1816 for hotels), guest capacity, environmental label.

The golden rule nobody tells you: third-party sources

This is the most common trap. Wikidata requires a verifiable, independent source for every claim. Your official website is enough for factual details (address, phone, opening hours), but for qualitative elements (reputation, awards), you need a third-party source: a press article, a tourism guide, an official database.

Without a third-party source, your item will get flagged for deletion. Our rule: we never create a Wikidata item for a brand with zero outside mentions. If a business doesn't have one yet, it's better to first get a mention in a local guide or media outlet before creating the item.

Step-by-step method (20 minutes per item)

1. Create a Wikidata account

Free sign-up at wikidata.org. Use a personal account (not a "business" account under your establishment's name — editing your own item is frowned upon).

2. Check that an item doesn't already exist

Search for the exact name plus the town. If an item already exists (often partial), complete it rather than creating a duplicate.

3. Create the item with the 6 base properties

Click "Create a new item." Fill in Label and Description, then add the properties one by one through the interface. Every value needs at least one reference.

4. Link the item from your site

On your site, add a sameAs in your JSON-LD schema pointing to the Wikidata item:

{
  "@context": "https://schema.org",
  "@type": "LodgingBusiness",
  "name": "Hôtel du Roc",
  "url": "https://hotel-du-roc04.fr",
  "sameAs": [
    "https://www.wikidata.org/wiki/Q..."
  ]
}

This sameAs is the bridge that tells AI models "this site and this Wikidata entity are the same thing." Without this line, the link is one-directional and far less effective.

5. Wait for indexing (1 to 4 weeks)

Wikidata takes a few days to propagate. AI models reindex at their own pace — expect 2 to 4 weeks for ChatGPT and Claude to pick up a freshly created item. Be patient.

Our real-world results from 5 client sites

We published clean Wikidata items for 5 establishments in the Verdon region (Castellane and the surrounding area): a hotel, two campsites, a spa, a rural gîte. Every item contains the 6 base properties plus a two-way sameAs from the site.

Results observed after 3 to 4 weeks:

• ChatGPT recognizes the establishments by name (a precise answer instead of "I don't have information on that").
• Perplexity cites the official sites in answers to local queries ("best hotels Castellane pool").
• Google's Knowledge Graph gets enriched in turn (Wikidata signals indirectly feed into Google).

Total cost: €0. Total time: half a day for all 5 items.

One question first: is your site already visible to AI?

Creating a Wikidata item without first checking that AI crawlers can actually read your site is like hanging a sign on a locked door. Livada Radar tests both in 30 seconds: bot access and the presence of an entity identity (Wikidata, sameAs). Free, no sign-up needed for the score.

→ Test my AI visibility for free

What's next?

A Wikidata item is a foundation, not an end goal. Turning entity recognition into regular AI citations also takes clean JSON-LD schema on your site, an llms.txt, and structured content that answers your customers' real questions. Livada SEO automates all of this on WordPress — including generating the Wikidata sameAs as soon as an item is detected.

Be found by Google. And by AI.

"Found by Google. Cited by AI." Wikidata is the most underrated lever for local AI visibility. Free, durable, and read by every model. If you run a local business and don't have an item yet, it's probably the best return on time you can get this month.