Your brand shares its name with another professional, another city, a different graphic variant — and every time, the AI has to guess which of the two you're looking for. Often it guesses wrong, or doesn't guess at all, and confuses you with someone else. Without a unique identifying code in the models' knowledge system, all the work you do on content and reputation rests on unstable foundations. Checking whether you have one takes ten minutes — and getting one, if you don't, is simpler than you think.
The question isn’t “is my brand on Wikidata?”. The question is “does it have a unique identifier that AI engines can resolve without ambiguity?”.
Let me explain it with a case that’s been rattling around my head for months. In Naples there are several historic pastry shops competing for the same imagery, plus a dozen artisan bakeries with overlapping names (“Forno Sorbillo”, “Forno Antico”, “Pasticceria Poppella”, “Pasticceria Capparelli”). If I ask ChatGPT “best historic pastry shop in Naples”, the model has to decide which entity to anchor the answer to. And it doesn’t do it by reading your site: it does it by looking for a unique identifier in the knowledge graph.
That identifier, today, is called the Wikidata QID. It’s the code that inherits the role of Google’s old Freebase ID (shut down in 2016 and migrated right into Wikidata). It’s a string like `Q12345` that tells the AI engine: “this entity is that one, not another one with the same name”. Without a QID, your brand is an ambiguity for an LLM to resolve. With a QID, it’s a resolved node.
In this article I’ll explain what really changes when a brand has a QID, the test I ran on the Neapolitan bakeries and pastry shops, and how to move concretely to get one.
What an AI engine sees when it searches for your entity
In the world of entity linking research — the discipline that studies how machines connect a name in text to a unique entity in the world — Wikidata is by now the central hub. Cedric Möller et al. (2021), in their survey on entity linking, map the entire landscape of datasets, and the picture is fairly clear.
“The identifiers follow either a human-readable form if available via Wikipedia or Wikidata or use the Wikidata QID”
Translated: when you need to tie an entity to a global identifier, the options on the table are Wikipedia, Wikidata, or the Wikidata QID directly. It’s not an industry opinion, it’s the operational default of the datasets on which AI models are trained and evaluated.
The consequence for your business is that if your brand doesn’t have a QID, the AI engine sees you as “one candidate entity among many”. It can confuse you with a namesake, it can merge you with another organization, it can simply discard you because the cost of disambiguation is too high. You don’t appear in the answer.
Why the QID sits upstream of all the work you’re already doing
In earlier articles in this series I told you about how AI models build the vector space in which concepts live and how Google and LLMs recognize an author’s authority. Two different processes, with an identical prerequisite: the engine first has to know who it’s talking about.
The QID is exactly that. It’s the primary key. Without a primary key, all the work on E-E-A-T, on schema markup, on well-structured content, hangs on an ambiguous text string. With the QID, every signal you emit anchors to a stable node in the knowledge graph and accumulates over time.
Cedric Möller et al. also describe how modern datasets are reannotating old corpora to anchor them precisely to Wikidata.
“The original KORE50 dataset focused on highly ambiguous sentences. All sentences were reannotated with DBpedia, Yago, Wikidata and Crunchbase entities.”
The message is clear: annotations are being redone to align them with Wikidata identifiers. Whoever doesn’t have a QID is out of the modern entity linking game.
Someone opens the Wikidata item, fills in only the name and description, and thinks they’re done.
The 10-minute test to find out if your brand has a QID
This is an entry-level check, a first step. The real analysis of the knowledge graph requires professional tools, but to figure out where you stand, start here.
- Go to Wikidata and search for your brand’s exact name.
- If an item appears with a `Q` code followed by numbers (e.g. `Q42`), you have a QID. Open it and check that the fields are populated: location, P31 (instance of), P17 (country), official website.
- If nothing appears, your brand doesn’t have a QID. It’s literally invisible at the entity linking level.
Second step: open ChatGPT or Perplexity and ask “who is [your brand name]”. If the answer is precise and cites data consistent with your company, there’s probably already an anchor (sometimes it comes through Wikipedia, sometimes through the Google Knowledge Graph). If the answer is generic or wrong — like “I don’t have specific information” or it confuses you with a namesake — the engine isn’t resolving you.
Third step: run the query for your sector (“best historic pastry shops in Naples”, “Neapolitan artisan bakeries with sourdough”). Look at who gets cited. The brands that appear in AI answers almost always have a QID or a Wikipedia entry behind them.
Add the Wikidata link as `sameAs` in the site’s `Organization` schema.
The test I ran on Neapolitan food
I took 15 brands from historic Campanian food — a mix of historic Naples pastry shops, artisan bakeries from the center and the province, a few businesses from Gragnano and Torre del Greco — and I cross-referenced two things: presence of a Wikidata QID and presence in AI answers to sector queries.
The pattern: 4 brands out of 15 had a populated QID. Of these 4, 3 regularly appeared in ChatGPT and Perplexity answers to the query “best historic pastry shops Naples” or variants. The other 11 brands without a QID appeared occasionally and almost always only when the query was very specific (containing the brand’s exact name, which is the least interesting case).
Then I ran a dirty check: I took 3 of the brands without a QID and asked directly “tell me about [brand name]”. In 2 cases out of 3 the engine confused the brand with a namesake from another region or returned inconsistent information. The classic signature of unresolved ambiguity.
Indicative test, not a peer-reviewed study: small sample, vertical sector, non-randomized queries. But the pattern is clear enough to stand on: the QID isn’t a magic factor, but its absence explains a good part of AI invisibility in sectors where namesake overlap is high — and traditional Italian food is full of namesakes.
The mistakes I see most often
When I talk about QIDs with SME clients, four patterns keep coming back.
Believing that having a Wikipedia page is enough on its own. Wikipedia helps but it’s a consequence, not the cause. The QID exists regardless of Wikipedia and in many cases it’s created first. Focusing only on “let’s write a Wikipedia page” means aiming at step 2 while skipping step 1.
Creating an empty QID. Someone opens the Wikidata item, fills in only the name and description, and thinks they’re done. A QID with 3 fields is noise. It needs to have P31 (what type of entity you are: bakery, pastry shop, food company), P17 (Italy), P131 (Naples/Campania), official website, year of foundation, and at least 2-3 verifiable external references.
Ignoring the link with other profiles. The QID needs to be connected — as an external identifier — to the Google Business profile, the company LinkedIn page, the official site. Without these hooks, the node stays isolated and other systems don’t resolve it.
Never referencing it in the markup. In your site’s `Organization` schema, the `sameAs` property can point right to your Wikidata URL. You can verify it with Google’s Rich Results Test: paste the homepage URL, look for the Organization block, check that `sameAs` includes the Wikidata link. If it’s not there, you’re leaving it on the table.
What to do concretely in the coming weeks
- Check whether you already have a QID on Wikidata. If so, open it and check that the key fields are populated with external references.
- If you don’t have one, create the entry respecting Wikidata’s notability criteria (cite external sources: local press, institutional mentions, gastronomic guides in the food case).
- Add the Wikidata link as `sameAs` in the site’s `Organization` schema.
- Connect the QID to the Google Business, LinkedIn profiles, and to the relevant sector databases (for food: Gambero Rosso, Slow Food, industry guides).
- After 4-6 weeks, rerun the query “who is [your brand]” on ChatGPT and Perplexity: the answer should become more precise.
- Compare with the 3-5 competitors that the AI cites in your sector: almost certainly they have populated QIDs. Study their fields.
The thread: why this matters for your visibility in AI answers
All the work you do to show up in answers from ChatGPT, Claude, Perplexity and Gemini rests on one prerequisite: the engine has to recognize you as a unique entity. Without a QID, you’re talking to systems that still don’t know who you are. You can have the best content in the world on twentieth-century Neapolitan pastry: if the LLM doesn’t resolve your entity, that content ends up anchored to someone else or to no one.
The QID is the base from which everything else starts — authority signals, markup, external links, citations. It’s the only way to capitalize over time on the work you’re doing: every new signal adds to a node that already exists, instead of dispersing into ambiguity.
In the next articles in this series I’ll explain how to correctly populate a Wikidata item for an SME brand, how to connect it to the other company identifiers, and how to recognize when the AI engine is resolving you well and when it isn’t.
