ChatGPT leans to chains, Gemini to independents
Same question, same city, two engines. In the US, ChatGPT cites 6.1 businesses per answer and 27% of its citations go to names found in three cities or more. Gemini cites 3.6, and 16%. In 9 trades out of 12, the gap points the same way.
6.1 / 3.6
businesses cited per answer, ChatGPT / Gemini
In this article
Short answer: ask ChatGPT and Gemini the same question and you do not just get different names, you get a different kind of business. Across 15 US trades, ChatGPT cites 6.1 businesses per answer and 27% of its citations go to names that come back in at least three cities. Gemini cites 3.6, and only 16% of its citations go to those names. For plumbers, ChatGPT names Roto-Rooter in 17 answers; Gemini, in none.
27% / 16%
of citations go to names found in 3 cities or more
9 of 12
trades where ChatGPT leans more toward those names
For an independent business, the two engines are not the same ground: one gives long lists where networks have their seats, the other short lists that stay local.
What we measured
The data is the Waseit US study of September 13, 2026. For this article we split the two engines across the 15 trades measured in all 13 cities (hotels excluded, since their names carry the city): 390 answers per engine, or 26 per trade (13 cities, 2 questions).
A name that travels is, as in the September 14 article, a name cited in at least three cities for the same trade, by either engine. It is not always a chain: unrelated businesses sometimes share a name. Directories are counted separately, and advice headings slipped into the lists are left out.
Fifteen trades, two engines
| Trade | Answers naming a business | Businesses cited per answer | Names that travel |
|---|---|---|---|
| Staffing agency | 100% / 92% | 8.0 / 5.0 | 87% / 54% |
| Accountant | 100% / 92% | 6.0 / 3.7 | 44% / 29% |
| Insurance broker | 100% / 96% | 7.4 / 6.0 | 36% / 41% |
| Driving school | 100% / 81% | 7.8 / 2.9 | 41% / 26% |
| Locksmith | 85% / 89% | 5.0 / 2.3 | 26% / 15% |
| Hair salon | 100% / 58% | 7.8 / 2.0 | 27% / 6% |
| Plumber | 100% / 85% | 6.4 / 2.8 | 28% / 3% |
| Auto repair shop | 100% / 96% | 7.6 / 4.8 | 26% / 6% |
| Real estate agent | 65% / 89% | 2.5 / 5.3 | 26% / 2% |
| Electrician | 92% / 62% | 5.8 / 2.2 | 11% / 2% |
| Wedding photographer | 100% / 96% | 7.1 / 3.8 | 7% / 9% |
| Marketing agency | 100% / 96% | 7.8 / 3.9 | 2% / 3% |
| Attorney | 58% / 89% | 2.7 / 5.4 | 0% / 0% |
| Dentist | 77% / 58% | 3.8 / 1.9 | 0% / 0% |
| HVAC contractor | 100% / 69% | 5.4 / 2.2 | 0% / 0% |
ChatGPT: long lists, with seats for the networks
ChatGPT names a business in 92% of its answers, Gemini in 83%. And when it names, ChatGPT gives about 70% more. In those long lists, the networks come back: Roto-Rooter in 17 ChatGPT answers about plumbers, zero for Gemini. Midas in 11 answers about auto repair, zero for Gemini. RE/MAX in 6 answers about real estate, zero for Gemini.
The real estate agent in Houston sums up the gap. To the same questions, ChatGPT cites Coldwell Banker Realty, Compass, eXp Realty and Keller Williams Realty, four national networks. Gemini cites Ashley Tilton, Beth Wolff, Laura Sweeney and The Amy Bernstein Team, local agents and a local team. Not one name in common.
Gemini: short lists, more local, and some silence
With Gemini, the share of names that travel drops to 3% for plumbers, 2% for real estate agents and electricians, 6% for auto repair shops and hair salons. Its lists are shorter, and what they hold is more local.
But Gemini stays silent more often on some trades: it names no business in 42% of its answers about dentists and hair salons, 38% about electricians, 31% about HVAC contractors. There, it answers with advice.
The exceptions, which matter just as much
Insurance brokers.
Here Gemini is the one leaning to networks (41% against 36%): it names Marsh McLennan in 16 answers, ChatGPT in 9.
Attorneys and real estate agents.
Here ChatGPT holds back: it names a business in only 58% and 65% of its answers, against 89% for Gemini, which cites twice as many.
Wedding photographers and marketing agencies.
Almost no name travels, on either engine.
What this measurement does not prove
Answer length.
We capped output at 600 tokens for ChatGPT and 1,400 for Gemini: 71 of 436 ChatGPT answers were cut off, mostly long lists. Its name counts are therefore a floor, and we do not compare text length.
What an app user sees.
The answers come from the APIs; the app can add a web search and other settings.
A deliberate preference.
We observe the shape of the lists, not the model's intent.
A perfect count.
Names are cleaned of advice headings; an error in that cleaning would weigh on both engines. And 26 answers per engine per trade is still a sample.
What it changes for you
Independent in a trade where networks come up
(plumbing, auto repair, locksmiths, hair): Gemini's lists are more local; ChatGPT's are longer but shared with national brands. Check both, they are not the same competitors.
Attorney or real estate agent
: ChatGPT names someone in only about six answers out of ten, Gemini in nearly nine. Start by reading what Gemini says about your city.
Dentist, hair salon, electrician
: Gemini names nobody in nearly two answers out of five. ChatGPT names someone far more often. That is where the list is decided first.
Ask your customers' question on both engines, several times: one engine's answer says nothing about the other. Our free audit does it for you and lists the names cited instead of you, engine by engine.