Appearing in AI answers for India and a second country at once, and why one reading cannot cover both

A business selling into India and one other country has to be visible in two different answers, and the two are built from almost entirely separate sets of sources. On 22 October 2026 we ran country audits for India and the United Kingdom on ChatGPT and Gemini, the same instrument on the same day. The India arm cited 1,149 distinct hostnames across 50 blind readings and the United Kingdom arm cited 1,023 across 32. Only 158 hostnames appeared in both, which is 13.8 per cent of the Indian set and 15.4 per cent of the British one. Both audits used country specific questions, so part of that difference is the wording and part is the market, and the practical consequence is the same either way. A single reading does not tell you where you stand in two countries, and work aimed at one answer does not land in the other by itself.

This page sets out what the two readings actually showed, and what a business with two markets should build.

The two countries side by side

Read the most used source row twice. In India the answer that two engines reached for most often was Google's own developer documentation. In the United Kingdom it was a small British agency's own website, in 20 of 32 readings, and the second and fifth most used sources were gov.uk at 15 readings and ico.org.uk at 9. India's largest government source was pib.gov.in at 7 readings. Two markets, one week, two completely different source landscapes, and only one of the two mechanisms is the same. The mechanism that holds in both is that over nine tenths of what these answers are built from is a page somebody published on their own domain.

The thing that fixes the market, and the thing that does not

This is the most expensive mistake available in multi country measurement and it is worth stating plainly. The market is fixed by the words in the question, not by a parameter on a tool. We measured this directly. A search provider accepts a country code, a language code and a location string, returns a success response for every value, and answers from somewhere else entirely. So a dashboard reporting your United Kingdom position because it set a country parameter is reporting a reading whose market was never fixed.

The working method is to write the country into the question text. Who are the best suppliers of this in the United Kingdom is a British reading. The same question with a country parameter attached and no country in the wording is a reading from nowhere. Every figure on this page comes from questions with the country named in them.

What to build for two markets

• One page per market for anything with a price, a regulation or a delivery promise in it. Not a translated page, a page written for that buyer, with that market's currency and that market's proof.
• Every currency in the HTML, on every request. This is the one we got wrong ourselves and it is worth copying the fix. Our pricing page detected the visitor's region from five signals, and a crawler sends none of them, so the detector fell through to its last line and served rupees to everybody who was not a browser. An assistant asked what this costs in the United States read our page and found rupee figures. The cause was not the default, which protects an Indian buyer whose phone ships with a British locale from seeing a ladder more than twice his own. The cause was that the other currencies were never in the document at all. Fetched as GPTBot after the fix, the visible text carries 15 dollar figures, 5 pound, 5 euro and 31 rupee, printed on every request whether or not anything about the visitor is known.
• Name the market in the text of the page. If the page is for Indian buyers, the word India belongs in it, along with the city, the currency and the registration. The market is fixed by words in a question and it is recognised by words on a page.
• Proof from near the buyer. A case from a company in their market answers a question a case from another continent cannot, which is why the two most used British sources after the agency itself were the government's own sites.
• The buyer's vocabulary, not yours. Indian English, British English and American English differ in the exact nouns a buyer types. One product can carry three names across three markets.

How to measure both markets without doubling the work

• Freeze one question set per market, with the country named in every question, and keep them separate files.
• Keep about two thirds of the questions equivalent across markets, so you can compare, and let one third be specific to each market's rules, currency and buying habits.
• Never name your own company in a question. A question containing a brand name primes the answer and the result cannot be compared with a blind one.
• Record three things per reading. Which businesses were named, whether you were one of them, and every source cited.
• Re run both sets unchanged after sixty days, so the second pass is a comparison rather than a new study.

The two figures to watch are different in each market. Where most readings name nobody, as in 31 of India's 50 and 20 of the United Kingdom's 32, the work is to publish the page that answers the question properly. Where a rival is named repeatedly, the work is to read the pages that answer cited and see what they carry that yours does not.

Questions about multi country AI visibility

Can one dashboard report my position in two countries

It can when each country is measured with its own question set and the country is written into the questions. What it cannot do is take one set of readings and split it by market afterwards, because the market was decided when the question was asked.

Do I need separate domains per country

No. What the measurement supports is a distinct page per market with that market's currency, wording and proof in the document on every request. A separate domain is a bigger decision driven by other things entirely.

Which engines do these numbers cover

ChatGPT and Gemini, two engines rather than five. Perplexity, Claude and Google AI Overviews were asked nothing in either country audit. Our product reads five surfaces and this dataset covers two, and those are separate statements.

Why were the two question banks not identical

Because each country audit asked about that country's own rules, currency, buying habits and city structure, which is where the commercially useful questions are. That is also why the source overlap figure is reported as a consequence of both the market and the wording rather than of the market alone.

Was either audit run in a language other than English

No. Both were in English with the country named in the wording. A language level reading is a separate measurement and neither dataset carries one.

How often should a two market reading be taken

Every sixty days on frozen question sets. Changing the questions and the date in the same pass gives you two variables and no finding.

The reading both markets share

One number came back almost identical in both countries, and it is the one to build on. 91.1 per cent of India's sources and 92.9 per cent of the United Kingdom's were pages on a company's own website, while review and listing directories took 0.7 per cent and 0.9 per cent. Two markets with 13.8 per cent of their sources in common agreed almost exactly on where the answers come from. That is the argument for publishing your own market specific pages rather than buying profiles in either country. AI Knows Us runs the per market reading above and names every source behind each answer, and the plan that reads all five surfaces is Growth, at Rs 11,799 a month in India or 299 dollars in the United States and the United Kingdom, read on our own pricing page on 22 October 2026. The benchmark across 597 readings in three markets publishes the method so the sixty day version above can be run without us.