AI Search: 68,000 AI Responses Reveal Fintech Visibility

Why Fintech Is Becoming the AI Search Laboratory

I watch 15 sites a week. All of them have the same problem.

They optimize their pages for Google. Titles, tags, internal linking. Then I ask a simple question in ChatGPT. « What’s the best budget tracking tool for a freelancer? » The answer cites three brands. Not theirs. Invisible.

Neil Patel’s study analyzed 68,000 AI responses to fintech questions. Concrete questions. « How do I send money to my relatives? » « What’s the best app for automatic savings? » The results show a shift. Ranking well on Google is no longer enough. The machine has to cite you.

Why fintech first? It’s a high-value sector. A misplaced AI recommendation costs thousands of euros. Trust is the product. Fintech brands were the first to want to measure their AI visibility precisely. A question about the best savings account, and the user opens an account online. A money transfer app recommendation, and they download it within the hour. The decision cycle is short.

Fintech questions are also very repetitive. Thousands of users ask the same variations. This creates a massive corpus where you can see which brands come up, where, and in which catégories. That’s what Neil Patel’s team did. They built a 100-point score. Three ingredients. Frequency, position, catégories. We’ll detail that right after.

How many e-commerce owners have this number today?

The AI Visibility Score: A 100-Point Score That Changes the Game

The score comes down to one idea. Nobody calculated it before.

The exact definition, according to Neil Patel:

Our AI Visibility Score is a 100-point measure combining how often a brand appears in AI answers, how early it’s mentioned, and how many catégories it shows up in.

Translation for an e-commerce owner. Frequency is the number of times your brand shows up across all responses. Position is the timing: first sentence, or end of the answer. Category diversity is the number of different topics you’re recommended for. One type of question, or ten.

Frequency. Position. Catégories.

Three dimensions. None of them is optional.

Concrete example. 1,000 AI responses to fintech questions. A budget management brand appears in 340. Its frequency is 34%. It’s cited in the first sentence of 280 responses: excellent average position. It shows up for budgeting, savings, investing, credit: strong category diversity. The 100-point score normalizes these three dimensions to compare brands against each other.

The raw data: 68,000 responses. Not 500. Not 2,000. 68,000. It’s starting to look like a market.

What I observe with my clients: a site can rank first on Google for « robot vacuum comparison » and be completely absent from ChatGPT responses for the same intent. Why? The model doesn’t read rankings. It reads content, entities, coherence. A site with ten pages cannibalizing each other on the same keyword sends no clear signal. The AI prefers to cite a site with a clear pillar page and well-defined product entities.

The 100-point score makes the diagnosis actionable. You know where you stand. You know what to improve. That’s what fintech teaches us: AI visibility is measured with the rigor of an advertising metric.

The Trust Triangle: DOSE and Early Repetition

This score isn’t just for reporting. It builds trust.

The DOSE framework, taught by Guillaume Attias at BMO Academy, explains how the human brain decides. System 1 is fast. It’s automatic, emotional. It doesn’t compare ten options. It connects what comes up often with what’s reliable. An AI recommendation doesn’t read like a comparison table. It reads like advice. And advice that puts a brand first, again and again, leaves a mark.

In an AI response, the early and repeated mention of a brand acts as algorithmic social proof. Your brain doesn’t read « this brand has a good credibility score ». It reads: this brand is named first, so I remember it. Cognitive fluency does the rest. What’s familiar seems true. What’s cited often seems safe.

Take a fintech question: « What’s the best app to track my subscriptions? » If the AI answers « Mint, YNAB, and Rocket Money are often recommended… », your System 1 records Mint first. It doesn’t matter that YNAB has more features. First place creates perceived authority. Repeated across dozens of questions, it strengthens the anchor. That’s the mechanism of top-of-mind awareness, applied to conversational responses.

Same thing in e-commerce. « What’s the best mattress for side sleepers? » If your brand is cited in the first sentence in 12 out of 20 responses, and that repeats across mattresses, bed frames, and bedding, your category authority climbs. The customer hasn’t opened your product page yet and already trusts you. The DOSE framework describes this exact moment: the decision happens before the analysis.

Why remember a brand cited in third position? The brain doesn’t calculate. It associates.

Early repetition builds memory. Memory leads to clicks. And clicks lead to sales.

Voici le résultat brut de l’audit e-commerce : la marque n’est citée que dans 7 réponses sur 40. La répartition parle d’elle-même — et elle est transposable à votre catalogue.

A Transferable E-commerce Case: 230 Pages, 40 Commercial Queries

An e-commerce client asked me for an AI visibility audit.

230 product pages. 40 commercial queries tested on ChatGPT, Perplexity, and Gemini.

What I saw with him: his brand showed up in 7 out of 40 responses. Never before the third sentence. Eight catégories of questions where he should have appeared naturally. He wasn’t there. The queries covered his products, their uses, their accessories. The responses cited his competitors. Three main competitors held the top spots.

Seven. Forty. Three.

We stopped producing new product pages. We restructured.

We rebuilt semantic cocoons: pillar pages, real questions pulled from customer support and reviews, product pages rewritten to answer conversational intents. Each category got a pillar page covering question variations. Product pages were linked with precise internal linking. Product, brand, and use entities were structured in Schema markup.

Three months later. The brand shows up in 19 out of 40 responses. Six times in first mention. Average order value climbed 21%.

An architecture, not luck.

The most interesting part: the pages didn’t change in volume. They changed in coherence. The site already had enough content. It lacked the structure that tells the AI which content to recommend for which question. Once the structure was in place, the mentions came. Without advertising. Without additional production.

The Classic Trap: Producing More When You Need to Restructure

Many e-commerce owners think AI visibility is won by publishing more content.

Mistake.

A catalog of 800 SKUs, no semantic structure. The owner produced 30 product pages a month. His brand appeared in no AI responses. None. Pages piled up, intents overlapped, the model no longer knew which page to recommend. So it always recommended the competitor.

We stopped production. We invested the next $8,000 in restructuring the site architecture and consolidating pillar pages.

+290% AI mentions on high-value commercial queries. Without advertising.

The real problem was architecture.

This trap comes back every week. The fear of not publishing pushes you to create pages that cannibalize each other. The AI no longer knows which page to recommend. So it recommends the competitor with a clear, unique, structured page. Overproduction dilutes the brand entity. Each extra page, without a precise link, weakens the overall signal.

Restructuring pays more than adding. Clarifying pays more than multiplying. One pillar page answering 15 question variations beats 15 repetitive product pages.

AI prefers coherence over volume.

Pour savoir où vous en êtes, on suit toujours le même protocole. Quatre étapes simples, applicables dès cette semaine sur votre propre catalogue.

Your Method to Calculate Your AI Visibility Score in 72 Hours

You want to know where you stand? Here’s the protocol we use in audits. Simple. Concrete.

• List 20 to 30 commercial questions your customers actually ask. Support tickets, reviews, Google Search Console, sales conversations.
• Query three AI engines: ChatGPT, Perplexity, Gemini. Ask each question once.
• For each response, note: is your brand cited? At what position? In how many different catégories does it appear?
• Calculate three ratios. Frequency: responses with your brand / total. Average position: 1 for first sentence, 2 for second, 3 for beyond. Diversity: number of catégories covered / catégories possible.
• Multiply by a simple coefficient, for example frequency × 40 + inverse position × 40 + diversity × 20. There’s your 100-point score.

A worked example. You test 30 queries. Your brand appears in 12 responses. Frequency: 40%. Of those 12 mentions, 8 are in the first sentence, 2 in the second, 2 in the third. Your average position is good, say 0.75 out of 1. You cover 4 catégories out of 6 possible, or 0.67 diversity. Your score approaches 40 × 0.4 + 40 × 0.75 + 20 × 0.67, or 16 + 30 + 13.4 = 59.4 out of 100. A measurable starting point. Three free engines are enough.

Then look at the pages that answer best. Strengthen them. Consolidate the ones that cannibalize each other. Create pillar pages. Structure product, brand, and use entities. Query the engines again after four weeks. The score moves as soon as the architecture becomes clear.

I’ll show you the pages. It’s more telling.

And the question that matters: what is the first sentence the AI says when someone talks about your market?

Stéphane Jambu

SEO & AI Engineer

I build growth systems / AI / Neuroscience | 650+ clients · 80 LinkedIn testimonials · 30 years of expertise · 15 years of systems running without me.

Follow on LinkedIn