AI doesn't cite you because your website is well built: it cites you because you're inside the specialized catalogs of your industry — Booking, Miodottore, category portals — that the models use as their primary source. For the AI, those who aren't there simply don't exist in that industry, regardless of the quality of what they do. Your less talented competitor who is listed in those catalogs gets recommended instead of you every single day. Figuring out which directories to claim — and how to do it so the AI truly recognizes you — takes less than fifteen minutes.
The brands that AI cites in your industry all have one thing in common: they appear in the 2-3 vertical directories that matter. If your industry directories don’t know you exist, neither does the AI.
It’s the first thing I check when a client tells me “I don’t show up in AI answers.” Before the website, before the content, before the knowledge graph. I look at whether the brand exists in the vertical catalogs of its industry. Nine times out of ten the answer is no, or it’s an empty profile with just a name and address.
In this article I explain why vertical directories have become the first place ChatGPT, Perplexity, Gemini and Claude draw from when they answer about a niche, and what you need to do to avoid being left out.
What I mean by a vertical directory
A vertical directory is an industry-specific catalog where the entities in that sector are listed, described and made comparable. It’s not a general-purpose directory like the Yellow Pages. It’s a database built for a precise niche, kept up to date by people who live and breathe that niche.
Booking for hotels. Avvo and Altalex for lawyers. Miodottore for doctors. TripAdvisor for restaurants. And, to stay with the example I’ll use today, IBS, Amazon Books, LibriPubblicati, IPL (Informatizzazione del Patrimonio Librario) and the SBN catalog for those in publishing.
Every industry has its 3-5 dominant directories. The point is that AI models, when they have to answer about a vertical, start from there. Not from the brand’s website. Not from a general-purpose blog. From the industry database they consider a clean, structured source.
Why AI trusts vertical directories
In the world of knowledge graph research, the documented mechanism is this: language models anchor the entities they know to nodes in structured graphs, and those nodes are populated by sources with high information density and low noise. Vertical directories are exactly that: standardized listings, consistent fields from one entity to the next, very little free-form text.
This leads to something concrete for your business: if you’re inside a vertical directory relevant to your industry, the model associates you with a clear category and a set of comparable attributes. If you’re not there, to the model you’re an uncertain entity — or simply don’t exist as a player in that niche.
I touched on this at a more technical level when I explained how embedding in vector space works: entities that inhabit well-defined contexts end up in dense regions of the model’s vector space. Vertical directories are among the densest and cleanest contexts that exist. Being there means having precise coordinates on the AI’s mental map.
The brand opens a profile on the vertical directory, fills in half the fields, uploads no photos or documents, and doesn’t update it for two years.
The comparison I ran: two independent publishers in Mantua
Let me tell you about a comparison I ran at the request of a client in publishing, because it’s the case that makes the effect most visible.
Two independent publishers in Mantua, both with a catalog of specialized nonfiction and cultural catalogs for institutions. Similar revenue, similar age, same average print run. I’ll call them Publisher A and Publisher B for confidentiality.
Publisher A is present on IBS with a complete publisher listing, has all its titles on Amazon Books with clean metadata (BISAC, ISBN, description, table of contents), is registered on LibriPubblicati with an up-to-date publishing profile, has SBN codes for every release and is in the IPL library catalog.
Publisher B has a well-made website, a carefully curated newsletter, publishes the same number of titles, but on IBS it shows up only with a name and city, on Amazon Books the metadata is incomplete (no BISAC, generic descriptions), it’s not on LibriPubblicati and hasn’t claimed SBN systematically.
I tried 12 queries on ChatGPT and Perplexity along the lines of “independent nonfiction publishers in Mantua,” “who publishes cultural catalogs in Lombardy,” “small Italian publishers of specialized nonfiction.” An indicative test, not a statistical study — small sample, high variability. Publisher A was cited in 9 answers out of 12. Publisher B in 1 answer out of 12, and in that one it had an attribution error on a title.
This leads to an operational conclusion that doesn’t depend on the single case: when the AI engine has to compile an industry list, it starts from the directories it considers reliable for that vertical. Publisher B’s well-built website is useful when someone is already searching for its name. It’s not useful for getting it into the set of candidates the model considers.
Identify the 3-5 dominant directories in your industry.
The 20-minute test on your niche
You don’t need an agency to run a first check. All you need is an hour and a bit of honesty when looking at the results.
Open ChatGPT or Perplexity. Write 5 queries along the lines of “best [category in my industry] in [my city or region],” “who are the main [type of player] in Italy in [niche],” “recommend [product/service] from small [industry] businesses in [area].”
Look at the answers. Note down the brands cited. Open a page and manually search for these brands on Google Trends (trends.google.com) to see whether they have significant search volume and which other terms get associated with them. Then look them up in the vertical directories of your industry: do they appear with a full profile? The correct category? Complete reviews or listings?
The decision threshold is binary: either the brands cited by the AI are in the same 3-5 vertical directories, or they’re not. In the vast majority of B2C and professional sectors I’ve looked at, the answer is: yes, they’re all there. And if you’re not there, you’re already cut out of the recommendation set.
The mistakes I see most often
In the projects I work on, the wrong patterns around this are always the same.
Profile created and abandoned. The brand opens a profile on the vertical directory, fills in half the fields, uploads no photos or documents, doesn’t update it for two years. To the AI it’s a zombie profile: it says “I exist” but doesn’t say “I’m alive.”
Wrong category at registration. An artisan pasta maker in Gragnano who registers as “general food products” instead of “artisan pasta maker from Campania” excludes itself from vertical queries. The category in the directory profile is the first filter the model applies.
General-purpose directories instead of vertical ones. Being on the Yellow Pages is almost useless if your industry already has its own specialized catalogs. AI models weigh a clean profile on a niche vertical directory far more heavily than a generic profile on a horizontal directory.
No consistency between website and directory. Different legal name, different address, different category. The model can’t connect the dots and treats you as two weak entities instead of one strong entity. I wrote about this when I explained why author entity recognition matters: the consistency of the signals around an entity is what makes it recognizable.
What to do concretely this week
This is a first step, not a complete analysis — that requires professional tools and work on the extended knowledge graph.
- Identify the 3-5 dominant directories in your industry. If you don’t know them, ask your best clients where they found you or where they look for suppliers like you.
- Open your profile on each one and run a completeness check: exact name, category, description, photos, contacts, reviews, any certifications or awards.
- Align the data between website and directory: same legal name, same category, same address, same phone number.
- Compare your profile with the 3-5 competitors the AI cites in your industry. The difference you see is the difference between being recommended and being invisible.
- Put a quarterly update on the calendar: new titles, new products, new awards, new reviews. Directories reward those who keep them alive.
The thread connecting to visibility in AI answers
Vertical directories aren’t a magic factor and aren’t enough on their own to get you into AI answers. But they’re the first signal the model looks at when deciding who to put in the candidate set for a niche query. Without that signal, all the work on the website and content is pushing against a closed door.
In the next articles in this series I’ll show you how to build your profile on Google Business Profile as a local entity, how to work on structured organization data, and how to bring your brand together on Wikidata so it becomes an official node in the knowledge graph.
