If your website says "we offer custom industrial solutions", the AI has already placed you in the wrong category — and when a buyer's specific question arrives looking for exactly what you do, the model doesn't even consider you. It's not a content problem: it's that a generic description excludes you from the results for the searches that really matter. Rewriting a few key texts with the industry and specialization stated clearly is often all it takes to get back on the radar of the right customers.
Brands that ChatGPT classifies in the right industry receive roughly ten times more relevant citations than those that stay generically categorized as “services”. Over the past six months I have observed the same pattern across more than thirty industrial food packaging companies: when the AI engine understands that you make food packaging machines instead of “industrial solutions”, citations in niche queries change by an order of magnitude. In this article I explain how you change classification, why it matters for your visibility in AI answers, and the test you can run this afternoon to find out where the model places you.
What the model really does when it “understands” your industry
In the world of applied NLP research, the mechanism generative models use to label a company within an industry is a form of zero-shot text classification: the model doesn’t consult a rigid taxonomy, it reads the available text (homepage, about, schema, third-party profiles, articles that cite you) and builds a probability that you belong to a certain category.
I don’t have a direct paper on industry vertical classification applied to commercial AI answers — nobody has yet published one with this angle. What exists is the general principle, abundantly documented in the literature on zero-shot classification: the category assigned depends on the density and consistency of the textual signals the model finds about your entity.
From this follows a consequence that became clear to me by observing clients in packaging: if your homepage says “we offer custom industrial solutions”, you are giving the model an ambiguous signal. If it says “we manufacture flowpack packaging machines for baked goods and food-grade plastic films”, you are giving a concentrated signal. In the first case you compete in the “services” melting pot, in the second in your real niche.
Why it sits upstream of everything you do for AI
In this series I have already written about how AI reads a text by turning it into vectors, where similar concepts sit close together in the numerical space — if you skipped that step, you’ll find the mechanism explained in the article on embeddings and vector space. Industry classification is the biggest projection the model makes of you in that space: it decides which region of the graph it places you in.
If you end up in the “professional services” region, you’re near lawyers, accountants and generic consultants. The model, when it receives the query “Italian manufacturers of food packaging machines”, searches in the industrial packaging region, not in services. If you’re not there, you don’t exist for that query. Even with impeccable E-E-A-T (I covered it in E-E-A-T for AI), it’s not enough: authority counts within your category, it doesn’t move you across categories.
“Since 1978 the X family has carried on a tradition of quality” is emotional and positions you nowhere.
The ten-minute test: where ChatGPT classifies you today
You can find out where you’re classified without professional tools. Open ChatGPT or Perplexity and try this sequence of prompts on your brand. I’ll take as an example a company from Reggio Emilia in food packaging, but adapt it to your case:
- “What industry does [brand name] operate in?”
- “List five Italian companies that make [your specific niche, e.g.: flowpack machines for food]. Is [brand name] among them?”
- “What are the main Italian manufacturers of [generic category, e.g.: industrial machinery]? Do you mention [brand name]?”
Read the answers with two binary thresholds in mind:
- Threshold 1: on the first question does the model use your specific category (food packaging, packaging machines) or a generic one (industrial services, B2B solutions)? If it’s generic, you’re in the melting pot.
- Threshold 2: on the second question, the one about your real niche, do you appear? If you don’t appear in your niche but you do appear in the generic category of the third prompt, the diagnosis is clear — the model knows you but has classified you wrongly.
It’s an entry-level check, not a complete analysis. The real analysis requires professional AI search monitoring tools and an audit of textual signals. But to find out whether you have a category problem, ten minutes are enough.
Rewrite the first sentence of your homepage with the supply chain made explicit.
The pattern I saw across thirty packaging companies
Over the past six months I have observed more than thirty Italian companies in industrial food packaging — manufacturers of packaging machines, food-grade plastic films, filling lines. It’s not a controlled study, it’s a longitudinal observation of my portfolio and of analyses done for prospects: a non-statistical sample but a consistent pattern.
The pattern is this. Companies whose homepage, meta description and Organization schema explicitly contained the supply chain (“food packaging”, “food-grade”, “food industry”) were cited by ChatGPT and Perplexity in industry queries like “Italian manufacturers of flowpack machines for the food sector”. Companies with generically industrial texts (“packaging solutions”, “technologies for industry”) were cited only in extremely broad queries (“packaging companies in Italy”) where competition is enormous and AI answers almost always pick the first three or four well-known names.
The difference in relevant citations — the ones coming from commercial queries, not from generic queries — was on the order of eight to twelve times, in favor of companies with explicit classification. It’s the figure behind the opening of this article. Take it as an indicative pattern: a tenfold increase in useful citations when you switch from “services” to “industrial food packaging”.
The mistakes I see most often
There are four patterns I find almost everywhere, even in companies with twenty to fifty million in revenue:
- Homepage written for the investor, not for the industry. Phrases like “strategic partner for industrial innovation” don’t contain a single industry signal. The AI model has no material to classify you.
- Generic Organization schema. Many manufacturers use `”@type”: “Organization”` without `industry` or without a technical description. Google’s Rich Results Test tells you whether the schema is valid, not whether it’s informative: judge for yourself whether the description field really explains what you do. You’ll find the test at search.google.com/test/rich-results.
- About pages based on storytelling without industry nomenclature. “Since 1978 the X family has carried on a tradition of quality” is emotional and positions you nowhere. You need at least one paragraph with the supply chain made explicit.
- Misaligned third-party profiles. LinkedIn, Google Business Profile, Wikidata: if three authoritative sources classify you differently (one “manufacturing”, one “industrial services”, one “food packaging”), the model struggles to consolidate. Go to business.google.com and to wikidata.org and check that the category is the same as on the homepage.
What can you actually do?
The intervention that works is more tedious than it seems: it consists of making your industry a consistent repetition on every textual surface the AI reads.
- Rewrite the first sentence of your homepage with the supply chain made explicit. Not “packaging solutions”, but “flowpack packaging machines and food-grade plastic films for the food industry”.
- Update the meta description with the same nomenclature. It must contain the specific industry, not the value promise.
- In the Organization schema, populate `description` with a sentence that contains industry + products + target market.
- Align LinkedIn, Google Business Profile and Wikidata on the same category. If you have a Wikidata entry, check the `instance of` and the `industry`.
- Compare your texts with those of the three to five competitors the AI cites in your niche when you ask “best Italian manufacturers of [your specialty]”. If they name the supply chain five times per page and you name it once, you know what’s missing.
It’s not a magic factor: industry classification won’t make you beat a competitor more authoritative than you. But it removes the first obstacle — being invisible in the right query — and puts you in the game. From there what counts is E-E-A-T, third-party mentions, and all the other signals I explained in the previous articles.
From the industry to visibility in AI answers
The thread of this series is always the same: showing up in AI answers when a potential customer asks a commercial question. Industry Vertical Classification is the point where the engine decides whether you’re a candidate for the question or not. If the answer is no, everything else isn’t even evaluated.
In the next articles of the series we’ll see how to consolidate the classification with finer signals: relationships between entities (suppliers, customers, trade associations), disambiguation when your brand name is ambiguous, and the role of co-occurrence patterns in the texts of the third parties that cite you. If you start from the wrong category, all those interventions build on unstable foundations. If you start from the right one, every subsequent signal adds up.
