If you search for your brand on Google and that box with logo, address and hours doesn't show up on the right, ChatGPT and Perplexity treat you as if you didn't exist in local queries — no matter how polished your website is. It's not a content problem: you're missing the primary identity that AI models consult first when someone searches for a business in your area. Competitors who have it get recommended in your place every single day. Building it is the starting point for everything else.
Open Google, search for your brand. Scroll to the Knowledge Panel on the side. Is it there? Is it complete? We bet it isn’t.
If the answer is “no” or “partly”, you’ve already found the first reason why ChatGPT, Perplexity and Gemini don’t cite you when someone asks for “best artisan jewelers in Vicenza”. It’s not a website problem, it’s not a content problem, it’s not a backlink problem. It’s a problem of a missing primary entity, and your primary entity for local queries is the Google Business Profile.
In the previous articles in the series on entities and the knowledge graph, I explained how AI engines build an identity record for every brand before deciding whether to cite it. For local businesses — jewelers, professional practices, pastry shops, boutique hotels — that record almost always starts from a single place: the GBP listing. Today I’ll explain why, and what to do about it concretely.
What it means to be a primary entity for an AI model
In the world of knowledge graph research the documented principle is simple: when an entity exists in multiple sources, the system picks one “canonical” source — the one with the richest data structure, the most frequent updates and verified metadata — and uses it as the foundation. The other sources become enrichment, not foundation.
It follows that for a local business the question isn’t “do I have a website?” but “what is my canonical source from the AI’s point of view?”. If you’re an artisan jeweler in Vicenza, the answer is almost always Google Business Profile. Not by magic: because it’s the only source where the system finds, in the same record, legal name + verified address + business category + hours + services + products + reviews with text + geotagged photos. No website, however well made, offers this level of structuring out of the box.
The operational consequence is direct: a GBP that’s 100% complete beats one that’s 60% complete in every AI-generated local answer. And “60% complete” is more or less the standard I see in Italian artisan SMEs.
Why GBP sits upstream of all the rest of your GEO work
You can have the best “about us” page in the world, you can have perfected author entity recognition for your founder, you can have structured your content as an inverted pyramid as I described in the articles on AI content structure. But if the user’s query contains a geographic modifier — “in Vicenza”, “near me”, “downtown” — the AI engine starts from the GBP listing and only afterward looks for confirmation on the website.
This is the same logic I showed you for E-E-A-T applied to AI: authority signals serve to confirm an entity that’s already recognized, not to create it from scratch. If the entity doesn’t exist in the local knowledge graph, everything else is enrichment of an empty container.
Generic single category: “Store” instead of “Artisan jeweler”.
The test you can run in 15 minutes
Go to Google Business Profile and check your account. I won’t tell you to “verify completeness” in the abstract — I’ll tell you exactly what to look at:
- Categories: one primary + at least 2-3 secondary. An artisan jeweler has at least “Jeweler”, “Goldsmith”, “Jewelry repair”, possibly “Watch store” if it sells watches.
- Services: a precise list. Not “repairs” but “necklace repair”, “ring resizing”, “stone setting”, “gold polishing”.
- Products: at least 10-15 products with photo, description, price or price range.
- Service area: specific municipalities (Vicenza, Altavilla Vicentina, Creazzo, Arcugnano), not a generic “Veneto”.
- FAQ/Questions: a “Questions and answers” section with at least 5 Q&As written by you (anticipate the questions, don’t wait).
- Posts: at least 1 post a week over the last 3 months.
- Photos: at least 30 photos, geotagged if possible, with team faces, interiors, products, the work process.
Binary threshold: if any one of these blocks is empty or below threshold, your primary entity is incomplete and the AI sees it.
Write 5-8 Q&As in the Questions and answers section, starting from the real questions customers ask you in the shop
The reverse engineering I did on local queries
I ran an indicative experiment — not a study, a test to understand the pattern. I asked Perplexity and ChatGPT (with browsing) ten local queries like “best artisan jewelers in Vicenza”, “historic goldsmiths in downtown Vicenza”, “where to get a ring repaired in Vicenza”, extended also to other SME sectors (pastry shops in Modena, tailors in Naples, boutique hotels on Lake Garda) to get a sample less biased by my own area.
Across 10 queries total, the brands cited in the answers had characteristics in common in 7 cases out of 10:
- GBP with at least 4 categories filled in
- more than 100 reviews with a reply from the owner
- a populated products/services section
- GBP posts published within the last month
- a website linked from the GBP listing with the same name+address (NAP consistency)
The brands that were NOT cited — even though they had better websites in some cases — had GBPs at 40-60% completeness, a single category, zero recent posts, reviews with no replies.
Limitations of the test: small sample, Italian queries, a single point in time. It’s not a controlled study. But the pattern is clear enough to warrant a serious audit of your profile. The real analysis requires professional AI position monitoring tools, but this entry-level test already tells you whether you’re in or out.
The mistakes I see most often
In the profiles I audit as a consultant, these are the recurring patterns that explain months of invisibility in local AI answers:
- Generic single category: “Store” instead of “Artisan jeweler”. The AI doesn’t know exactly what you sell and excludes you from specific queries.
- Empty services or duplicated from the website: copy-pasting the website menu doesn’t work. GBP wants short, singular service names designed for semantic matching.
- Zero posts for over 6 months: the system reads the profile as “dormant” and lowers its weight in generated local results.
- All photos look the same, no process: the AI (and the user) look for visual authenticity. For an artisan business, photos of the workbench, the tools, the master goldsmith at work weigh more than 50 glossy product shots.
- Reviews with no replies: the signal of activity and customer care is missing. Replying to every review — positive and negative — is the bare minimum.
What to do concretely this week
Take 3 hours, no more, and work in this order:
- Open GBP and bring the categories from 1 to 3-4 (primary + secondary ones consistent with what you actually do)
- Populate products and services with specific names, not brochure-style descriptions
- Add a service area with individual municipalities, not regions
- Write 5-8 Q&As in the Questions and answers section, starting from the real questions customers ask you in the shop
- Reply to ALL reviews from the last 12 months
- Plan 1 post a week (promotions, new arrivals, behind the scenes) for the next 3 months
- Compare your profile with the 3-5 jewelers that Perplexity cites when you ask “best artisan jewelers in Vicenza”: see what they have that you don’t
This is the work that underpins everything else in your AI-answer visibility strategy. If the complete primary entity isn’t there, any content you publish afterward is noise that the system struggles to connect to you.
How it ties into the rest of the work on entities and the knowledge graph
Google Business Profile is the starting point, not the finish line. In the next articles in this series we’ll see how to strengthen the entity with external signals — LocalBusiness schema markup on the website, NAP consistency across industry directories, presence on Wikidata for brands that reach a certain level of notoriety. But if GBP is incomplete, working on those layers is like building the second floor before finishing the first.
The thread is always the same: AI answers local queries by drawing from well-structured entities. If you want to show up in those answers — and for a business with a physical store, not showing up means losing concrete chunks of revenue — the GBP listing is the first job site, not the last.
