Two assistants, the same question, no name in common
We put the same 1,320 questions to ChatGPT and Gemini in four countries. Four times out of five, the two answers name no business in common.
79.2%
of questions where the two assistants named no business in common
In this article
Short answer: being named by one assistant tells you almost nothing about the other. We put the same 1,320 questions to ChatGPT and Gemini in four countries, and in 79.2% of them the two answers shared no business at all. Same wording, same city, same trade, asked the same day. Two lists of about six names each, and nothing in common.
1,320
questions put to both assistants, word for word, in four countries
0 of 24
plumber questions in the United States where the two shared even one name
Being visible in one assistant is not being visible. It is being visible in one assistant.
What was measured
Waseit’s four national studies hold the raw answers: France, collected on 30 July 2026, the United Kingdom on 31 July, Brazil on 6 August and the United States on 13 September. Each study asks two questions in 13 cities across about twenty trades: what are the best {trade} in {city}, and list the most recommended {trade} in {city} and explain why. Both assistants receive the same sentence, checked word for word. We kept the questions where each named at least three real businesses, which leaves 1,320 pairs out of 1,800, and compared the lists two ways: exact matching, then a deliberately generous one where a shared word, or one name contained in another, counts as agreement.
| Country | Questions compared | No name in common | Still nothing, matching generously |
|---|---|---|---|
| United States | 357 | 82.6% | 38.9% |
| United Kingdom | 343 | 74.9% | 42.6% |
| France | 241 | 70.1% | 19.5% |
| Brazil | 379 | 85.5% | 41.7% |
Where they agree, and where they never do
The disagreement is not uniform. It collapses in trades run by large branded networks, and it is total in trades that are purely local.
| Trade | Country | Questions with at least one shared name |
|---|---|---|
| Staffing agency | United States | 67% (16/24) |
| Estate agent | United Kingdom | 68% (15/22) |
| Recruitment agency | United Kingdom | 61% (14/23) |
| Hotel | United Kingdom | 45% (9/20) |
| Dentist | United States | 5% (1/22) |
| Wedding photographer | United States | 4% (1/24) |
| Attorney | United States | 0% (0/20) |
| Plumber | United States | 0% (0/24) |
| HVAC contractor | United States | 0% (0/19) |
| Dentist | United Kingdom | 0% (0/13) |
A staffing agency question lands on a handful of national brands, and both assistants reach for them. A plumber question has no national brand to reach for, so each assistant builds its own list, and the two never meet. The more local the trade, the less they agree.
What it means if you are the business
An audit that looks at one assistant measures one assistant. Being named in ChatGPT is worth having, and it says close to nothing about what Gemini answers the same customer asking the same question. For a local trade these are two separate markets, and the second is not a copy of the first.
What this does not prove
- It does not say which one is right. Neither list was checked against the businesses that actually serve those cities.
- It does not measure how many customers ask. The studies measure answers, not traffic or demand.
- The four studies were collected on different dates, between 30 July and 13 September 2026, so the comparison across countries is not a same-day comparison.
- Perplexity was left out: it is missing from the United States study, and including it would compare countries on different engines.
- The names come from an automatic extraction of the answers. We removed 476 distinct phrases that were criteria rather than businesses, such as “Strong Track Record”, and dropped the 480 questions out of 1,800 where one assistant ended with fewer than three usable names.
- One test, once, is not a year of testing. Each assistant answered each question a single time.