Ask ChatGPT for the best accounting software for a small business from one country, then again from another, and the two answers share 36% of the products they name, on average. Ask twice from the same country and they share 78%.1
You asked an engine your category question from head office, saw three competitors and not yourself, and the result is about to go into a board update as your AI visibility. It describes one country.
Highlights
Read together, they say a visibility number measured from one country describes that country. How many people use AI, the engine they open and, for a local product, the shortlist itself all change at the border.
Empact Partners does this measuring for a living. We are a go-to-market consultancy embedded in B2B software companies’ own marketing teams.
Our Generative Engine Optimization (GEO) workstream gets brands recommended when buyers ask AI engines what to buy, and for a partner selling into several countries its first output is that number, per market.
AI Use Runs Highest in Small, Wealthy Countries, and the US Ranks 21st
Start with who is asking at all. Microsoft’s AI Economy Institute counts the share of people aged 15 to 64 who used a generative AI product in the quarter, from its own device telemetry adjusted for population and internet access. Its September 2026 report puts the country that built the engines nowhere near the top:2
The first three are small, wealthy and well connected, and none of them has a large home market. A working-age adult in Ireland is half again as likely to be using AI as one in the US, which is the first thing a rollout plan built in the US gets wrong.
OpenAI’s own data counts ChatGPT messages per person rather than people, and it puts the US lower still. Its ranking of 147 countries for the second quarter of 2026:5
Europe’s official statistics, a third ruler that counts any use in the last three months among people aged 16 to 74, agree on the shape: Norway 56.3%, Denmark 48.4%, the EU as a whole 32.7%, Italy 19.9% and Romania 17.8%.9
Size and intensity are different maps. India ranks 109th per person and is still ChatGPT’s second-largest market, with 100 million weekly users by February 2026, by OpenAI’s chief executive’s own count.10
China is missing from OpenAI’s ranking altogether and has more than 700 million generative AI users of its own, over half the population.11 Microsoft reads China from device telemetry and expects its next, wider measure to raise China’s figure the most.2
.png)
The reasons the sources give are mostly about the floor under the technology. Microsoft names gaps in electricity, connectivity, digital skills and local-language support as what holds the Global South back.2
Its early adopters there lean on AI to learn. Self-directed learning takes a 29.2% larger share of consumer Copilot conversations in the South than it does globally.2
So the US is a large AI market and a middling AI country, and a plan that rolls AI visibility work out from the US reaches the markets where use runs highest last.
Most People Meet AI Search Inside Google Before They Open a Chatbot
Using AI and searching with it are different acts, and the second mostly happens without anyone choosing an AI tool. The Reuters Institute surveyed about 2,000 people in each of six countries in June and July 2025 and asked about both:3
France is the control group nobody planned. The one country without Google’s AI Overviews had the lowest exposure, 17 points below the next country. In most markets the door to AI search is a feature Google switches on, and national exposure follows Google’s rollout calendar more closely than anyone’s appetite for chatbots.
Weekly AI tool use moved unevenly. Most of the six countries roughly doubled in a year, while the US went from 31% to 36%.3 Across the six, weekly ChatGPT use reached 22%, twice Gemini’s 11%.3 The engines’ own counts, and Pew’s, put the scale on it:
In most countries, then, the first AI answer a buyer reads sits on a Google results page. A low chatbot number is not low AI exposure, and Google’s AI surfaces get measured as a channel of their own in every market where they have launched.
ChatGPT Still Sends Most AI Clicks, and Gemini Takes More Every Quarter
When an AI answer does send a visitor somewhere, the engine behind it is still mostly ChatGPT. StatCounter counts the clicks AI chatbots send to the more than a million websites carrying its code, and in September 2026 the split looked like this:4
| Clicks from AI chatbots, September 2026 | ChatGPT | Gemini | Copilot | Perplexity |
|---|---|---|---|---|
| Worldwide | 80.48% | 10.91% | 1.79% | 5.85% |
| United States | 75.17% | 13.26% | 6.64% | 3.02% |
| United Kingdom | 80.02% | 10.14% | 5.61% | 2.7% |
| Germany | 79.27% | 11.06% | 3.08% | 5.15% |
| Brazil | 76.63% | 17.95% | 2.17% | 2.14% |
| India | 77.25% | 16.04% | 1.71% | 3.62% |
ChatGPT’s three clicks in four hold in every market. The engine in second place is where countries differ: Gemini’s share in Brazil is almost twice its share in the UK, Copilot’s in the US is more than three times its worldwide share, and Perplexity does better in Germany than in any of the other four markets.
The trend runs one way. Gemini’s worldwide share of AI chatbot clicks went from 2.31% to 8.65% between April 2025 and March 2026, while ChatGPT’s slid from 84.21% to 78.16%.15 By September 2026 Gemini stood at 10.91%.4
Similarweb’s panel, which counts visits to the AI sites rather than clicks out of them, tells a sharper version over June 2025 to May 2026:16
Similarweb’s visits and StatCounter’s clicks are two different rulers, so neither divides into the other. Side by side, Gemini holds about a quarter of the visits and about a tenth of the outbound clicks, which means ChatGPT still sends most of the traffic a website will see from AI.
The device question has a blunt answer, which is the app. ChatGPT became the fastest mobile app ever to reach 1 billion monthly users, in three years.17
Win ChatGPT for the clicks, and pick the second engine you measure by country. Gemini is second everywhere and closest in Brazil and India, Copilot carries more weight in the US than anywhere else, and Perplexity more in Germany than in the other four markets.
Why People Ask Changes by Country, and So Does When
Exposure and engines decide where a question lands. Habit decides whether it gets asked.
OpenAI publishes how a sample of 300,000 consumer ChatGPT messages a month splits by topic and by work, country by country. Looking something up, in its taxonomy, covers questions for specific facts, product research and recipes. For the five markets, the June 2026 cells read:5
| Share of consumer ChatGPT messages, June 2026 | Looking something up | Making or reading media | Work-related |
|---|---|---|---|
| Germany | 22.8% | 6.7% | 25.5% |
| United Kingdom | 21% | 6.6% | 32.3% |
| United States | 20.7% | 6.9% | 31% |
| India | 17.2% | 9.8% | 34.4% |
| Brazil | 15% | 11.9% | 35.2% |
| Worldwide | 18.8% | 7.8% | 30.4% |
Germany asks and Brazil makes. In Germany more than one ChatGPT message in five looks something up, the closest thing in the data to a search. Brazil looks things up least of the five markets and makes or reads media most. For a software company the first column is the one that matters, because it is the pool a category question sits in.
The last column answers when. Worldwide, the work-related share of messages on consumer plans fell from 51.3% to 30.4% between July 2024 and June 2026, so most of the asking now happens outside work.5 Europe’s statistics point the same way: 25.5% of EU adults use generative AI for private purposes, against 15.4% for work.9
Work is where people do rather than ask. At work, OpenAI finds, they are more than twice as likely to use ChatGPT to complete a task or create something.18 The askers are also getting older, which matters to anyone selling to a budget holder:
Wherever someone asked people why they use it, looking things up came first:
How many category questions an engine ever sees in a country is set by local habit before any answer is given. Weight markets by how much of their AI use is asking, and headcount second.
The Same Question Gets a Different Shortlist Where the Product Is Local
Everything so far describes the asker. The part a marketing team can act on is the answer, and nobody publishes whether it changes when the asker moves, so we measured it.
On 20 October 2026 we asked ChatGPT, Google AI Mode and Gemini 20 software-buying questions in English from the US, the UK, Germany, Brazil and India, three times each, 900 answers in all. We held the language still so the country was the only thing that moved, and left out the categories our current partners sell in.1 What came back:
The run says something narrower, and more useful, than “AI answers are local”. For a horizontal tool the list barely moves: two answers from different countries share almost as many names as two from the same one, and most of what differs is the churn you would see asking twice at home.
Where the product has to comply with local tax, invoicing or payment rules, the list is rebuilt at the border. Accounting on ChatGPT is the clearest case. Asked from Germany it added DATEV, Lexware and sevDesk, from Brazil Bling, Conta Azul and Omie, from India BUSY, Tally and Vyapar, and from the UK FreeAgent and Sage, each named in that country only.1
.png)
The engines disagree about what location means. ChatGPT followed the asker’s country. Google AI Mode and Gemini followed the question’s language: asked about tax software in English from Germany, Brazil and India, both named TurboTax, a US filing product, in 9 of 9 answers each.1
The citations split the same way. Asked from Germany, 9.7% of ChatGPT’s citations were German domains, against 1.7% for Google AI Mode and 0.2% for Gemini.1 Asked from the UK, where the language matches, 14.2% of Google AI Mode’s citations were British.1
Per-market lists are the part of SEO we read as Search Everywhere Optimization: every market a buyer searches in is its own surface, so before a partner launches in a country we run its category questions from that country, on each engine, and compare the list with home. A buyer typing in German or Portuguese adds AI search localization on top.
One Brand’s AI Visibility Moves 13 Points Between Its Best and Worst Market
Our run asked identical questions. Real buyers do not, and real tracking measures each market on its own questions, usually in its own language.
For that view we worked with Qvery, our sister company, which asks every tracked question daily on ChatGPT and Google AI Mode and records whether the brand comes up. It is ours, so weigh it accordingly, and test these figures against your own tracking before you take our word for them.
We took every brand Qvery tracked in more than one country between 1 July and 23 August 2026: nine brands, 38 brand-and-market pairs, 74,921 answers.6
The gap mixes the country, the language and the question set, which is the honest version of what a team sees: nobody tracks Germany with American questions.
The same mix is why one global visibility figure is an average of markets that move separately, and why Google AI Mode needs its own line in any report, since it named these brands more often almost everywhere.
That per-market baseline is the first thing a GEO engagement with Empact Partners produces: the questions buyers in each market ask, run daily on the engines they use there, with the gap to home written down before anyone spends money closing it.
The Click That Survives AI Answers Is Smaller, and It Arrives as a Brand Name
Visibility is the first half. The second is what an answer does to the visit. Ahrefs measured 300,000 keywords from December 2023 to December 2025 and found the top-ranking page’s click-through rate 58% lower when an AI Overview appears.7 Pew tracked the browsing of 900 US adults through March 2025:19
Google’s own business is fine, which is the part SEO plans tend to miss. Search and other revenue grew 17% in the second quarter of 2026, and Google says its AI features now send billions of clicks to websites every week.12
Ranking still feeds the answer. Across the 922 queries in Qvery’s study of AI citations against Google’s top ten, the higher a domain ranked on Google, the likelier an engine was to cite it for the same query.8
The click that does arrive has changed shape. After ChatGPT started turning brand names in its answers into links in May 2026, the share of its referrals landing on a homepage rose from roughly 26–29% to about 62–63%, in Similarweb’s data.16 A buyer who arrives from an AI answer arrives looking for you by name.
Keep ranking, because the engines read what ranks, and judge the AI channel where it lands, in branded and direct visits per market.
What Gets a Brand Named Is Written by Other People, Market by Market
If the engine reads other people’s pages, the work is making sure those pages exist in each market. GEO runs on one formula for it, GEO = UGC + Mentions: the reviews, threads and comments buyers post about a product, plus the pages other publishers write about it.
Which pages count depends on the engine. Qvery’s research on mentions in SaaS answers found listicles were 8.62% of ChatGPT’s citations against 2.50% of Google AI Mode’s, and Reddit the third most-cited domain on SaaS questions, at 1.57%.20 We build both halves for the market the buyers are in, which is not always the market the partner is in:
On the mentions half, Existing Article Outreach is our method for getting a partner into the roundups the engines already cite, and the per-market version picks the roundups each country’s answers cite rather than the ones that rank at home.
On the UGC half, our Reddit marketing runs from accounts with a real history in the communities each market’s buyers read, because a first post carrying a product link is the one a moderator removes.
In practice, working with Empact Partners on GEO starts with an audit and a per-market baseline, then runs two workstreams at once, usually mentions and community, with a named consultant inside your channels every week. It moves in quarters, and taking AI share of voice in another country takes longest.
What comes next is already in the data. Similarweb’s 2026 landscape report counts an ad in roughly 26% of ChatGPT responses.16 AI Mode went global inside a year and passed a billion monthly users,12 and ChatGPT reports 1.2 billion weekly users.13
For a software company selling across borders, the next AI visibility budget belongs in the market with the highest per-person AI use where its category is local, not in its largest market. What our evidence cannot settle is how far a buyer’s own language moves your list on top of that, because our run held it still.
If you are about to report one AI visibility number for a company that sells in several countries, book a call, and we will work out together whether this is a job for us.
Sources
- Empact Partners, the country run: 900 answers from ChatGPT (web search on), Google AI Mode and Gemini to 20 software-buying questions in English, asked three times each from the US, the UK, Germany, Brazil and India, collected through DataForSEO on 20 October 2026.
- Microsoft AI Economy Institute, Global AI Diffusion Q2 2026 Trends and Insights, September 2026.
- Reuters Institute for the Study of Journalism, Generative AI and News Report 2025: How People Think About AI’s Role in Journalism and Society, Felix Simon, Rasmus Kleis Nielsen and Richard Fletcher, October 2025.
- StatCounter Global Stats, AI Chatbot Market Share, worldwide and country pages, September 2026, read 20 October 2026.
- OpenAI, OpenAI Signals v2.0: data download and data dictionary, August 2026, CC BY 4.0; June 2026 cells and second-quarter 2026 per-capita ranks.
- Qvery, our sister company: each tracked brand’s visibility per market and per engine for the nine brands it tracked in two or more countries, 38 brand-and-market pairs and 74,921 ChatGPT and Google AI Mode answers, 1 July to 23 August 2026, read from Qvery’s analytics on 20 October 2026.
- Ahrefs, Update: AI Overviews Reduce Clicks by 58%, Ryan Law, February 2026.
- Qvery, AI Engine Citations vs Google Organic SERPs, March 2026.
- Eurostat, Individuals: use of generative AI tools (isoc_ai_iaiu), 2025 data, updated June 2026.
- Sam Altman, The Times of India, India has a head start on AI. Now it needs these 3 things to make it work, February 2026.
- China Internet Network Information Center (CNNIC), Generative AI Application Development Report (2026), September 2026.
- Alphabet, Q2 2026 earnings call: Remarks from our CEO, July 2026.
- OpenAI, DevDay 2026 Recap, September 2026.
- Pew Research Center, Americans and AI 2026: Chatbots, Smart Devices and Views on Impact, June 2026.
- StatCounter, Google Gemini overtakes Perplexity to become second largest source of AI chatbot referrals to websites, April 2026.
- Similarweb, AI Search Stats 2026: Market Share, Referral, and Citation Data, Maayan Zohar Basteker, July 2026.
- Sensor Tower, State of AI 2026, June 2026.
- OpenAI, From asking to doing: How the world is putting ChatGPT to work, August 2026.
- Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results, July 2025.
- Qvery, Mentions Are the New Backlinks for SaaS, June 2026.
- Empact Partners partnership record: a three-month Reddit program for a travel marketplace partner, as published in the partner’s case study, read 20 October 2026.
- Empact Partners partnership record: a marketplace partner’s international expansion, as published in the partner’s case study, read 20 October 2026.
