The AI cites you, but you're fourth or fifth in a list — and whoever reads that answer rarely makes it all the way down. Being there isn't enough: the first brand cited gets an attention the fourth one will never receive, exactly like what happens with the first result on Google. Monitoring where you appear in AI lists and knowing what pushes your brand higher are two things that make a huge difference in the number of customers who then actually contact you.
Being in the AI list isn’t enough. Are you first, fifth or last? In AI CTR, position matters as much as in SEO — maybe more.
Let me explain why. When a user asks ChatGPT “best clinical analysis center in Crema”, the model doesn’t reply with ten blue links. It replies with a list of three or four names, in order. The first one is read, evaluated, often clicked. The fourth gets skipped. The fifth, when it exists, is practically furniture.
In the articles of the series on measuring AI visibility I showed you how to count mentions and how to tell a useful citation from a decorative one. Now let’s take a step forward: tracking which position you’re named in. Because position, inside the answer of a generative model, isn’t a cosmetic detail. It’s the factor that decides whether the user stops at you or moves on to the competitor below.
What “AI Recommendation Position” means
The AI Recommendation Position is the ordinal position of your brand within the list the AI model returns in response to a commercial query. First, second, third, or relegated to the bottom behind the phrase “others include…”.
I’m not talking about the position in the trailing citations (the linked sources). I’m talking about the body of the answer: where it names you, and where it places you relative to the others.
On this point I don’t have an academic paper to cite you. In the world of LLM research, the documented mechanism concerns two adjacent things: position bias in prompts (models weight the tokens near the beginning or end of the context more heavily) and the primacy effect in ranked answers, inherited from classic SEO literature where the first organic result absorbs most of the clicks.
From this it follows that the explicit deduction holds for your business: if the user reads an AI list top to bottom and tends to stop at the first names — exactly as they did (and do) on Google’s SERPs — then appearing in first position is worth 3-5 times more than appearing in fourth. It’s not a lab-verified figure. It’s an estimate consistent with human reading behavior and with what I’ve been observing on clients for months.
Why position matters more than the mention
In the first part of this series I kept telling you that the first goal is to be there: if the AI never names you, you’re invisible. True. But once you’re there, the problem immediately becomes another one: where?
Think of a biomedical outpatient clinic in Crema, the kind that does blood draws, clinical analyses, ultrasounds and specialist visits for the Crema area. Let’s say it has been around for twenty years, has 4.7 stars on Google, two locations between Crema and Bagnolo Cremasco. The average user no longer searches “analysis center Crema” on Google. They open Perplexity or ChatGPT and type: “where can I get a thyroid ultrasound in Crema with a fast turnaround”.
The AI replies with three names. If the clinic comes out first, it wins the appointment. If it comes out third, the user has already read the first two, weighed distance and hours, and probably won’t scroll further. Same city, same clinical quality, same prices — the only difference is the order of appearance.
This connects the discussion directly to what I told you in Author Entity Recognition and in Backlinks as a citation proxy: the authority signals that trigger the mention are often the same ones that decide the position within the mention.
Celebrating the mention and ignoring the order.
The test you can run yourself in 20 minutes
I’ll give it to you as a repeatable procedure, no paid tools.
Open ChatGPT, Perplexity, Gemini and Claude in four tabs. Prepare a list of 8-10 commercial queries from your industry. For a biomedical clinic in Crema, things like these work:
- “where to get blood tests in Crema without an appointment”
- “best center for thyroid ultrasound in the province of Cremona”
- “biomedical clinic Crema open on Saturday”
- “hormone tests Crema report in 24 hours”
For each query, note on a sheet:
- Absolute position: 1, 2, 3, 4, “among others”, or absent
- AI engine that answered
- Competitors cited before and after you
You don’t need an app. A Google Sheet with three columns is enough: query, engine, position. Repeat the same query 48 hours apart to check stability — AI models vary, but for local commercial queries the rankings tend to be fairly stable in the short term.
Simple, ternary decision thresholds:
- You come out first or second on high-intent queries → you’re winning, the work now is defensive
- You come out third or fourth → you’re seen but not chosen, you need to push on authority signals
- You come out “among others” or absent → you’re in emergency territory, it’s not a position problem but an existence one
This is an entry-level check, I’ll tell you straight. The real analysis, across dozens of queries and tracked over time, requires professional AI rank tracking tools. But the first signal you get this way, in half an hour, for free.
Build your own query grid (10-15 phrases a real customer would type in your industry) and track them every two weeks across at least three AI engines
What I’ve seen in 6 months on real clients
Over the last six months I’ve continuously observed about forty Italian clients — including two medical-diagnostic clinics in Lombardy, one right in the Crema area — tracking the position weekly across roughly a dozen commercial queries for each.
The pattern that emerges, and I’m reporting it with all the caveats (my own observation, not a controlled study, small sample, AI engines that change behavior from week to week), is this:
- The brands that start in fourth place and climb to first are the ones that, in the preceding six months, worked on three specific things: a complete Google Business Profile with updated FAQs, service pages with direct answers in the opening (the classic inverted pyramid), and citations in third-party sources in their own industry (trade associations, healthcare portals, local outlets)
- The brands that drop from first to third position almost always neglected freshness: pages not updated in 18+ months, few recent reviews, zero new external mentions in the period
- The brands that stay stable in a low position (“among others”) share one symptom: a shallow presence of the brand as an entity, no clear disambiguation on Wikidata or knowledge graph, authors/doctors not recognizable as individual entities
The Crema clinic I was telling you about went from “not cited” to “stable second place” on 6 queries out of 10 in about four months, working precisely on these three fronts. It’s not magic, it’s consistency between signals.
I’ll restate the caveat: it’s an observed pattern, not a law. But 40+ clients over 6 months is enough to stop calling it “a fluke”.
The mistakes I notice most
Four patterns that recur in the clients who come to me frustrated because “I’m mentioned but they don’t click on me”.
Celebrating the mention and ignoring the order. The classic “look, ChatGPT names us!” screenshot sent to the work chat. Nice. But if you’re fifth out of five, the user never reached you.
Tracking a single query, a single time. A query sees you first today, fourth tomorrow. The one-off data point says nothing. You need a grid of stable queries, monitored for weeks.
Confusing position in the body with position in the sources. Some AI engines show the links of the consulted sources at the bottom. Being “the sixth linked source” is not the same thing as “being named first in the answer”. The second matters much more: the user reads the answer, doesn’t always click on the sources.
Optimizing for Google and ignoring AI engines. The biomedical clinic that ranks first on Google for “analysis center Crema” can easily be fifth on Perplexity for the same intent. The signals overlap but don’t coincide. They’re parallel worlds, not the same world.
What to do concretely starting Monday?
Three practical actions, in order of priority:
- Build your own query grid (10-15 phrases a real customer would type in your industry) and track them every two weeks across at least three AI engines
- Compare your average position with that of the 3-5 competitors the AI cites before you. Go to their pages, look at who their authors are, their cited sources, their reviews
- Open Google’s Rich Results Test, paste the URL of your main services page, check that at least the `Organization` and `LocalBusiness` schemas come out (for a healthcare clinic, `MedicalBusiness` too). Without these basic signals, your brand’s disambiguation starts off limping
The thread that holds this series together
The whole series on measuring visibility in AI answers revolves around one idea: you can’t improve what you don’t measure, and you can’t measure well if you stop at “I’m there / I’m not there”. The AI Recommendation Position is the first layer of depth: where I am, relative to whom, on which queries.
In the next articles of this series I’ll tell you how to measure the stability over time of an AI position, how to build a DIY AI rank tracker with a spreadsheet, and how to do a structured comparison with the competitors who come out ahead of you.
Position, on its own, isn’t a magic factor — it’s not enough to sustain a business. But it’s the metric that translates AI visibility into real conversions, and ignoring it means leaving on the table the difference between a first place that works for you and a fifth that nobody sees.
