Run a search for your industry on Google and your name never appears in the AI Overview area at the top of the page. It's not a problem with the site or the content: for Gemini you don't yet exist as a recognizable company in its internal knowledge base, the one it uses to decide which names to cite automatically. Competitors who were added to it years ago get cited every time, but getting in is possible even starting from scratch.
Search a query from your industry on Google.
Is there an AI Overview at the top?
Are you cited in the list of 3-5 sources? No?
Here’s why you’re losing traffic.
If you run an architecture firm specialized in restoring medieval villages in Cosenza, try “architecture firms historic villages Calabria” or “old town restoration Calabria architect”. If you see a box generated by Gemini that cites three firms from the North — and yours isn’t there — the problem is almost never the site. The problem is that, as an entity, for Gemini you don’t exist enough.
Let me explain why and what you can do about it in the next 10 minutes.
What “knowing an entity” means for Gemini
The other AI engines (ChatGPT, Claude, Perplexity), when they receive a query, do more or less the same thing: they retrieve documents from the web, read them, and compose an answer. Gemini starts one step earlier.
Gemini is Google’s model. And Google, for over ten years, has maintained a structured map of the world: the Knowledge Graph. It’s not a copy of the web: it’s a database of entities (people, companies, places, concepts) with verified attributes and relationships. When you search “Renzo Piano” and see the panel on the right with photos, dates, works, and links — that’s the Knowledge Graph speaking.
In the world of research on retrieval-augmented generation, the documented mechanism is clear: models that have access to a structured knowledge base respond more accurately and with fewer hallucinations than those that read only free text. Google has stated it in black and white in its public guidelines for Gemini in Search: the Search Generative Experience rests on the web index and the Knowledge Graph together.
From this it follows that, for Gemini, you are not just the URL of your site. You are — or are not — a node in the graph. And if you aren’t, every other signal (content, backlinks, E-E-A-T) passes through a much narrower bottleneck.
Why this changes the rules of the game compared to the other AI engines
In the previous articles in this series I explained how tokenization works inside the models, and how Author Entity Recognition lets the AI recognize who’s speaking. Google’s Knowledge Graph sits upstream of all of this for a practical reason.
On ChatGPT or Perplexity you can win visibility through well-structured content, citations, and authority signals. They are engines that read the web and synthesize it. On Gemini there’s an extra step: before reading the web, it consults the graph. If your entity in the graph is complete — with the correct category, geographic location, official site, relationships with other known entities — the model has an anchored reference on which to build the answer. If instead you’re a black hole, Gemini struggles to understand who you’re talking about even when it reads your site.
The practical consequence is simple: a competitor with a mediocre site but a clean entity in the KG can regularly beat a competitor with a better site but invisible to the graph. It’s not fair or unfair, it’s just how the system is built.
Many sites have an “Organization” schema but with empty or inconsistent fields (a different address across the schema, Google Business, and contact details).
The field observation: 20 firms monitored for 6 months
Over the past six months I monitored 20 Italian professional firms of similar size — architects, engineers, building restoration firms — spread across the North, Center, and South. For each one I tracked two things: the presence of a Knowledge panel on the right side of Google for brand queries, and the frequency of citation in Gemini’s AI Overviews across 12 industry queries (village restoration, green building, conservative restoration).
The pattern that emerged is clear, even though the sample isn’t large and should be taken as indicative.
Firms with a visible KG panel — therefore with an entity recognized by Google — appeared in the AI Overviews on average for 4 queries out of 12. Those without a KG panel appeared on average for fewer than 1 query out of 12. The difference was explained neither by site size, nor by the number of backlinks, nor by content quality (I checked a sample). It was explained by the presence of the entity in the graph.
An indicative test, not a controlled study. But six months of observation on similar professionals tell a story consistent with what Google itself states about its system.
Create a Wikidata entry or, if one exists, enrich it with secondary sources (magazines, industry catalogs, local outlets that have cited you)
The test you can run in 10 minutes
First thing: find out whether you exist as an entity for Google. Go to Google and search for your firm’s name plus the city: “Studio Architetti Rossi Cosenza”. See whether a Knowledge Panel appears on the right (a panel with photos, address, hours, possible logos). If it appears, you’re recognized. If it doesn’t appear, you’re still outside the graph in your category.
Second thing: check how Google sees you in terms of structured data. Open Google’s Rich Results Test, paste in your home page URL, and look at the results. You should see at least one “Organization” or “LocalBusiness” schema recognized without errors. If you don’t see it, your site isn’t telling Google who you are in machine language.
Third thing: check Wikidata. Go to Wikidata and search for your firm’s name. Wikidata is the public backbone of the Knowledge Graph: if you’re here, with a well-compiled entry linked to official sources, you’ve taken a huge step toward entering Google’s graph.
Fourth thing: Google Business Profile. Make sure you have a profile on Google Business Profile that’s 100% complete, with the correct category (“architect” and not “agency”), description, photos, and recent reviews. Gemini uses these signals as a geographic and category anchor.
Decision thresholds: if you pass 3 or 4 of these 4 checks, you’re in a good position. If you pass 1 or 2, you have an enormous margin for improvement. If you pass none of them, Gemini is foreign territory for you.
The mistakes I see most often
Thinking the site is enough. The site is a necessary but not sufficient condition. Without an entity in the graph, the site is like a shop with no sign, in an alley with no name on the map.
Confusing traditional SEO with entity-based visibility. Ranking first on the query “architect Cosenza” in classic search doesn’t guarantee being cited in the AI Overview for the same query. They’re two different engines inside the same Google.
Schema markup copied badly. Many sites have an “Organization” schema but with empty or inconsistent fields (a different address across the schema, Google Business, and contact details). For the graph, this inconsistency counts for less than having nothing at all.
Ignoring Wikidata because “it’s an encyclopedia”. Wikidata accepts professional entities as long as they’re well referenced. A firm with 20 years of history, projects published in industry magazines, and mentions in local outlets has the right to an entry.
What to do in concrete terms
- Open the Rich Results Test and fix the Organization/LocalBusiness schema on the home page
- Complete the Google Business Profile 100% with a precise category
- Create a Wikidata entry or, if one exists, enrich it with secondary sources (magazines, industry catalogs, local outlets that have cited you)
- Make name, address, phone, and site consistent across every online presence (round up all the industry directories)
- Compare yourself with the 3-5 competitors Gemini cites in your industry: do they have a Knowledge Panel? Wikidata? a clean schema? Copy the structure, not the content
This is a first-level operational list. The real analysis — entity disambiguation, linking to parent entities, managing aliases and variants — requires professional tools and time.
Why this gets you into the AI answers
The thread running through all the articles in this series is the same: showing up in AI answers isn’t luck, it’s having the right signals in the right places. For Gemini, the most important place is the Knowledge Graph. Without you in that graph, every effort on content risks being a shout in an empty room. With you in the graph, every piece of content you publish has a hook to latch onto.
It’s not a magic factor and it’s not enough on its own — Gemini also weighs content freshness, source quality, and editorial authority signals. But it’s the foundational layer. In the next articles in the series I explain how to build visibility on ChatGPT and Perplexity (where the rules are different) and how to coordinate your presence across multiple AI engines without duplicating work.
