AI Search Traffic Patterns: why your GEO strategy must be tailored market by market
Summarize this article with AI
87.6 million AI visits analyzed: what Aleyda Solis’s study reveals
I come across Aleyda Solis’s study. 87.6 million AI clicks analyzed. Ten markets. The US, UK, Spain, France, Germany, Italy, the Netherlands, Australia, Mexico, and Brazil. March 2026.
Not a vague aggregate. A pixel-by-pixel snapshot of the traffic that conversational AIs send to websites.
The reflexive move in SEO is to hunt for constants. The study tells us the opposite. AI traffic is not a single thing. It shifts shape, concentration, and destination depending on country and vertical.
Two observations jumped out at me.
First observation: the concentration of AI clicks has nothing to do with one vertical versus another. The study counts the number of domains needed to capture 50% then 99% of clicks. In e-commerce, the median is 5 domains for half the traffic. In finance, 17 domains. In travel, 47 domains.
Translation: in e-commerce, if you’re not one of the 5 dominant domains, your chances of being clicked are slim unless you operate inside those domains. In travel, the door is wider.
Second observation: AI traffic often lands on local infrastructure, not global defaults. Amazon dominates, but its local version: Amazon.it, Amazon.fr, Amazon.de. In Italy, 46.2% of e-commerce AI clicks go to Amazon.it. Add Temu and you cross the 50% mark. In Germany, to reach 50%, 4 domains suffice, and among them, Otto and MediaMarkt—purely German brands.
The salience effect is brutal. Citational AIs surface brands already anchored in a market’s digital ecosystem. It’s a strong bias, and it varies from country to country.
This is not a nuance. It’s a deployment rule.
E-commerce: 2 or 3 domains can monopolize half your AI opportunities
Look at the data market by market:
| Market | Domains for 50% of e-commerce AI clicks |
|---|---|
| Italy | 2 |
| USA | 3 |
| Brazil | 3 |
| Germany | 4 |
| France | 5 |
| Mexico | 5 |
| United Kingdom | 6 |
| Spain | 7 |
| Australia | 9 |
| Netherlands | 11 |
These numbers are unambiguous. In e-commerce, concentration is extreme. The top 10 captures an average of 63.9% of AI clicks on the vertical. The first domain, on average, 26.9%.
I observe the same thing with my clients.
An e-commerce operator selling running shoes calls me. He’s pushing « best running shoes 2026 » content in English. Six months of production. Zero AI clicks. We analyze the landscape with Aleyda’s data: the US market is locked down by 3 domains, including Amazon.com and Nike. We won’t dislodge them.
We pivot.
I apply an inside-out strategy: instead of attacking generic queries, we optimize product pages on Amazon.fr, Cdiscount, and the own store with local semantic architecture. Each destination page is structured to answer the precise intent of the market—currency, delivery times, local shoe size variants. +210% AI traffic in 3 months. Primary source: ChatGPT and Perplexity.
The lesson? In e-commerce, if you’re not domain #1, #2, or #3, don’t fight the platforms. Invest in them.
Finance: trust is not decreed, it is localized
On this vertical, the median is 17 domains for 50% of clicks. Less concentrated than e-commerce, but the pattern is deceptive.
Look at Mexico: only 8 domains capture half the finance AI traffic. And not just any: BBVA México, Banorte, local institutions. In Brazil, 18 domains, led by Bradesco and Itaú. In Germany, 16, dominated by Sparkasse and Commerzbank.
The narrative of trust-led finance is real, but we read it wrong. Trust, for conversational AIs, is not measured solely by brand size. It’s measured by depth of local anchoring: a .de domain, mentions in German financial media, national comparison tools.
A French insurance broker solicits me for the German market. His site is .com, his citations are on Anglo-Saxon portals. No local trust signal. Worse: his pages don’t answer the reimbursement tables typical of the German market.
I build a dedicated semantic cluster.
We create pages like « Krankenversicherung Vergleich » with comparative data in the expected format. We’ll get mentions on portals like Finanztip. We shift the technical mesh toward a dedicated .de extension. In 5 months, AI traffic from Germany jumps +340%.
Does your trust policy go beyond simple customer reviews? For AIs, trust is a technical fact. A domain, a structure, local signals.
Travel: the AI funnel fragments discovery, concentrates booking
This is perhaps the most surprising vertical. 47 median domains to capture 50% of global AI clicks. But as soon as we observe the mesh by country, the gap explodes.
In the UK, it takes 129 domains to reach 50%. In France, 80. In Spain, 47. In Australia, 32. In Italy, 32. Travel is radically fragmented at the discovery layer.
This means one thing: AIs address queries like « what to visit in Andalusia » or « best season for the Maldives » to a myriad of local sites. But at the booking layer, the landscape tightens around comparison tools and national agencies.
A cruise operator based in Southeast Asia—a confidential client—contacts me. Their AI traffic is nearly zero outside their domestic market. We deploy a strategy of inspirational, localized content by market: itineraries departing from London, from Sydney, from Frankfurt. Each page adapted: currency, timezone, departure airports.
Result: +185% citations in English-language AI responses (UK, Australia) in 4 months. And clicks follow.
The mechanic is simple. In the dream phase, the AI distributes broadly. If your content is the only one speaking the market’s language—prices in pounds sterling, GMT schedules—you take the spot.
The DOSE framework for cracking AI citation logic market by market
These data give us the « what. » The vital question becomes: how do you position yourself in these contrasting traffic flows?
I rely on a framework I apply to each semantic cluster: the DOSE framework, created by Guillaume Attias within BMO Academy. It structures four levers that act directly on salience effects in GEO.
D for Differentiation. On an e-commerce market where 2 domains capture 50% of AI clicks, your content must stand out. Not another comparison guide. A niche angle, local, unfindable elsewhere. I delivered +820% AI traffic for a German auto parts site by creating 47 clusters differentiated by model and engine type.
O for Objectives. I determine precisely the volume of capturable queries on each market, leveraging data like Aleyda’s. No scatter.
S for Silhouette. Page structure must signal to the AI what it discusses, for whom, in which country. Localized schema markup, comparison tables, links to local sources. This is the technical layer of the salience effect.
E for Engagement. A page citing local studies, showing photos of real travelers from Lyon, embedding forms in euros. The AI detects real relevance, not a shell.
For the cruise site, I built one DOSE per market: differentiation on unusual itineraries, targets at 1,200 queries per country, silhouette with city-by-city departure points, engagement with local travel logs.
AI traffic followed. No ads.
3 steps to prioritize your GEO investments from market patterns
I’ll summarize my method.
Step 1: diagnose concentration in your vertical by market. Lean on Aleyda Solis’s data or your own analysis. How many domains capture 50% of AI traffic in your sector on your target country? Fewer than 5, platform orientation. More than 40, content orientation.
Step 2: identify domains cited by AIs. Not just clicks: list the sources ChatGPT, Perplexity, or Gemini use for answers on your keywords. These are your competitors in the citation layer, even if they don’t compete with you in classic SEO.
Step 3: build one DOSE cluster per market-vertical pair. For each priority market, a dedicated semantic cluster answering local signals, with page structure designed for AI.
An e-commerce client brought me 3 markets: France, Germany, Brazil. Diagnosis: concentration 5, 4, 3. We focused 70% of effort on optimizing local marketplaces (Fnac, Otto, Mercado Livre) and 30% on localized comparison content clusters. +152% global AI traffic in 6 months.
The key is not to cover everything. It’s knowing which markets to ignore because the battle is already lost.
Is your vertical closer to Italy’s 2 domains or the UK’s 129?
Live GEO audit: your pages analyzed in depth
In 40 minutes, I take your two priority markets, analyze your competitors’ pages already capturing AI traffic, and show you how to surpass them with a DOSE cluster. No sales pitch, just the mechanics.
Book a strategic call — 45 minFrequently Asked Questions
What is the salience effect in GEO?
It’s the tendency of conversational AIs to more easily cite brands already well-integrated into the local digital ecosystem. A site with a .fr domain, mentions in French media, and pages in euros will have a major advantage for queries issued in France. This bias is measurable and structural.
Are Aleyda Solis’s study data exploitable for my sector?
Yes, if you operate in e-commerce, finance, or travel across the 10 markets analyzed. For other sectors, concentration principles still hold: first measure the number of domains cited by AIs on your market to adapt your strategy.
Does the DOSE framework require training?
It’s taught within BMO Academy by Guillaume Attias. Implementation means structuring each destination page around the four levers. My clients benefit from a ready-made cluster, but you can train yourself to apply it independently.
Should we prioritize AI optimization or classic SEO?
The two reinforce each other. Signals that improve classic search ranking (local authority, semantic mesh, speed) also improve AI response presence. The reverse is true: a site well-cited by AIs gains in visibility. I build clusters for dual performance.
What are the first success indicators in GEO?
Track your domain appearing in AI response sources (not just clicks), growth in citations without links, and traffic evolution from AI platforms. An increase in indexed pages on DOSE clusters is also a good signal.
