If you bring your partner or your board a report on AI visibility full of technical metrics, the response you get is a polite silence — and the budget stays right where it is. The problem isn't the strategy: it's the language. The people who decide the money want to know whether the company shows up when a customer asks the AI, not read an index from zero to a hundred. There's a way to present the situation in terms that convince whoever holds the purse strings, and it works even if they've never heard of AI optimization.
AI knows more about your market than a consultant charging 10k a month. Ask ChatGPT what the emerging trends in your sector are, who is growing, what customers say about competing companies. It’s free market intelligence, and almost nobody uses it. The same is true in reverse: AI tells your potential customers who you are, or it doesn’t. The question is what level you’re at today on that scale, and how you explain to whoever puts up the money where you want to be in the next 12 months.
Let me explain the maturity model I use when I have to present the state of AI visibility to an entrepreneur or an operating partner. Six levels (from 0 to 5), binary thresholds, no jargon. It works because an entrepreneur doesn’t want to know the salience score of an entity (the importance of a specific entity): they want to know whether their company shows up in AI answers when a customer looks for it, and how far away they are from getting there.
Why you need a maturity model for AI visibility
In the previous articles in this series I talked to you about how AI engines think, about authority, about the knowledge graph, about content structure. All operational stuff. But if you walk into a meeting room with your partner or with the person running your sales team and tell them “we need to work on our implicit reference weight,” you lose them within two minutes.
The maturity model serves exactly this purpose: translating the state of AI visibility into a scale that anyone, even someone who doesn’t know what an LLM is, can understand in thirty seconds. It’s the same mechanism used in security with CMMI, or in digital marketing with marketing automation maturity models. Nothing new as a framework. What’s new is applying it to presence in the answers of ChatGPT, Perplexity, Claude and Gemini.
In the world of maturity model research there is not, to date, a consolidated paper that defines a standard for AI visibility. It follows that the model I’m proposing is a deduction: it comes from combining two principles documented elsewhere (the scale of organizational maturity models and the citation mechanisms of AI engines that I described in the articles on authority and credibility for AI), translated into an operational framework. I present it as such: it’s not a standard, it’s a working tool.
The six levels (0-5), explained the way I describe them to an entrepreneur
The model I use has six states, from 0 to 5. I’ll describe them with the question that lets you position yourself in thirty seconds.
Level 0 — Not monitored. You’ve never opened ChatGPT or Perplexity to search for your brand or your sector’s queries. You don’t know whether they cite you, you don’t know who they cite among your competitors. It’s the starting state of most Italian SMEs today. It’s not a fault, it’s the zero point.
Level 1 — Monitored, invisible. You’ve run the queries, you’ve realized you never show up. You know the problem exists. You know who shows up instead of you. You have a benchmark, but your brand isn’t in the answer set of the AI engines.
Level 2 — Sporadic mentions. Across ten queries relevant to your business, the brand comes up once or twice. When it does, it’s often in a marginal position (last on the list, mentioned in passing). It’s the signal that something about you has reached the training or the retrieval system, but without weight.
Level 3 — Mentioned regularly. Across ten relevant queries, the brand comes up six or seven times. Not always as the first choice, but consistently in the set. You’ve become “one of the names” the model considers when asked about something in your sector.
Level 4 — Among the top 3 recommended. When the user asks for “the best X in Y,” you come up among the first three named. It’s the position that corresponds more or less to the first page of Google in classic SEO, but with a multiplier effect: AI tends to summarize around a few names.
Level 5 — Default answer. When the conversation turns to your sector or your category, your brand is the first thing the model names, even without a specific question. It’s rare, it’s almost always the preserve of established brands or very vertical niches, and it isn’t a realistic goal for most companies.
The rule I always tell clients: the goal isn’t to reach 5. The goal is to move up one level in 6-12 months. From 1 to 2 is a job of basic authority. From 2 to 3 is a job of topical coverage and citation. From 3 to 4 is a job of reputation and links from sources the AI recognizes as authoritative — see what I wrote you about the backlink as a citation proxy.
Measuring on a single AI engine only.
The test you can run in twenty minutes to position yourself
Open ChatGPT, Perplexity, Claude and Gemini in four tabs. Prepare ten queries one of your ideal customers might ask. Not queries about your brand name: queries about the category. “Best producers of X in region Y,” “who do you recommend for Z,” “what are the most reliable brands for W.”
For each of the ten queries, note on a sheet whether your brand shows up and in what position. Do the same thing for the three competitors that come to mind first. It takes you about an hour. At the end you have a forty-cell matrix per brand. Count the appearances, divide by ten, and you have a rough percentage that maps to the maturity model fairly directly: 0% you’re level 1, 10-20% level 2, 50-70% level 3, over 70% level 4.
It’s an entry-level test, I’ll say it clearly. It doesn’t replace a serious analysis done with professional tools that run hundreds of queries automatically, across multiple markets and in multiple languages. But to say in the company “we’re at level 2, we want to reach 3 by year-end” it’s more than enough.
Second: for every competitor that shows up more than you, open three of the sources the AI cites and read what they say sets you apart.
The case I worked on: a Taggiasca olive producer in Dolcedo
For a few months I’ve been working with a producer of Taggiasca extra-virgin olive oil with a mill in the hinterland of Dolcedo (province of Imperia, the historic heart of the Taggiasca growing area). A family business, 80,000 bottles a year, mixed distribution (direct sales at the mill, e-commerce, a network of delicatessens in northern Italy). When we first spoke, the owner had a precise question: “my competitors show up on ChatGPT, I don’t. What should I do?”
The first thing I proposed: use AI as a competitor intelligence tool, even before working on his own visibility. For two weeks we ran systematic queries on Perplexity and ChatGPT like “best Taggiasca olive oil producers in Liguria,” “artisan Ligurian mills to visit,” “differences between Taggiasca and other Ligurian cultivars,” “who sells quality Ligurian extra-virgin oil online.” For each query we noted: who gets cited, from which sources, with what attributes (organic, historic mill, awards won, Slow Food presidium).
The pattern that emerged was clear: the three brands the AI cited most often had three things in common. A Wikipedia page or a Wikidata entry for the brand or the mill. Recurring presence on recognized online gastronomic guides (Gambero Rosso, Slow Food, trade magazines). Product pages on the website with a long description, the history of the cultivar, certifications declared in a structured way. Our producer only had the third thing, and not even done very well.
The client’s starting level, on the model: 1, monitored and invisible. We knew where we stood, we knew who to cite as a benchmark, we knew which levers were missing. Honest limitation of the test: a sample of 40 queries total, not a statistical study. A fairly clear pattern, though, to prioritize the next six months of work: work on the entity (see the piece on the Google Knowledge Graph entry) and citations from sector guides.
The mistakes I see most often when a maturity model is applied
Confusing the AI visibility level with revenue. I’ve seen companies at level 4 with low revenue and companies at level 1 with high revenue. The model measures presence in AI answers, not the health of the business. They’re two separate dimensions, even if over the medium term they influence each other.
Wanting to skip levels. From 1 to 4 in three months can’t be done. AI engines update slowly, authority is built over time, citations from recognized sources can’t be bought. Promising whoever decides the budget a three-level jump in a few months is the best way to lose the client soon after.
Measuring on a single AI engine only. If you run the test only on ChatGPT and ignore Perplexity, Gemini and Claude you risk positioning yourself in a distorted way. Each one has a different retrieval and training logic. The real level is the weighted average of the four.
Measuring only once. The level changes. A new article published by an authoritative site can move you up a notch in two weeks. A guide that stops citing you can move you down. The model should be reapplied every three or four months, not just once.
What to do Monday morning, if you’re level 0
Three concrete actions in order.
First: dedicate two hours to the ten queries about your sector across the four main AI engines, noting your presence and that of your competitors. Get out of level 0.
Second: for every competitor that shows up more than you, open three of the sources the AI cites and read what they say sets you apart. You have a map of the levers.
Third: bring the result (the scale from 0 to 5, where you are, where the three competitors are) to whoever decides the budget. This is the conversation the maturity model makes possible.
In this series on measuring AI visibility I’ll tell you in the next articles how to structure a continuous monitoring dashboard, how to estimate the economic value of a level jump, and how to differentiate investments across authority, entity and topical-coverage levers. The scale from 0 to 5 is the starting point: it gives you and the people working with you in the company a shared language to talk about visibility in AI answers without reducing it to technical slides nobody will read.
