
Half the recruiters AI names are national networks
In 18,662 recommendations from ChatGPT and Gemini, one recruiter in two is a national network, one accountant in five, and no dentist or restaurant.
1 in 2
names given for a recruiter are a national network (709 of 1,429)
In this article
Short answer: it depends on the trade. Ask ChatGPT or Gemini for a recruiter and one name in two is a national network (709 of 1,429). For an accountant, one in five; for a plumber, an electrician or a locksmith, 3%; for a dentist, a restaurant or a wedding photographer, none of 2,997. Across 18,662 business recommendations in four countries, at least 9.3% go to a network named in five or more of the 13 cities we asked about. The rest goes to independents, or to names that look local and are not quite: 360 recommendations are a trade and a city glued together ("Houston Driving School", "Plombier Paris"), and 299 of them come from ChatGPT.
0 of 2,997
names given for a dentist, a restaurant or bakery, or a wedding photographer are a network
360
recommendations are just a trade and a city ("Houston Driving School"); 299 come from ChatGPT
In some trades the AI's shortlist is a list of brands. In others, a brand never makes it.
What was measured
Our four studies, re-read on October 15, 2026: buyer questions put to ChatGPT (gpt-4o-mini) and Gemini through their APIs, without web search, in 13 cities in each of the United States (collected September 13, 2026), the United Kingdom (July 31), France (July 30) and Brazil (August 6), for 18 to 20 trades per country. 3,600 answers. A recommendation is one distinct business name in one answer: 19,419 of them, with the same name cleaning as our recent editions. 757 were not businesses: 610 headings read as names ("Especialização", "Get Multiple Quotes") and 147 platforms or bodies (property portals, Which?, the Brazilian bar). That leaves 18,662. A name is a network candidate when it is recommended, for the same trade in the same country, in at least five of the 13 cities. 343 names qualify; we read every one and sorted them: 146 networks we can name with certainty (Hays, Orpi, Midas, Roto-Rooter), 35 names made of the trade and the city, 62 names that are not a network we know, 100 that are not businesses. In doubt, a name went to "not a known network", never to "network". Names seen in three or four cities (2,181 recommendations) were not read: the network shares below are a floor.
| Trades | Recommendations | National networks | Share | Trade + city names | Same name, 5+ cities |
|---|---|---|---|---|---|
| Recruitment and staffing agencies | 1,429 | 709 | 49.6% | 0 | 61 |
| Accountants | 993 | 222 | 22.4% | 7 | 0 |
| Insurance brokers | 1,173 | 227 | 19.4% | 10 | 0 |
| Estate and letting agents | 1,447 | 214 | 14.8% | 16 | 0 |
| Driving schools | 1,040 | 97 | 9.3% | 77 | 65 |
| Garages and auto repair | 1,250 | 81 | 6.5% | 69 | 46 |
| Hotels | 1,104 | 52 | 4.7% | 0 | 0 |
| Plumbers, electricians, locksmiths, heating, air conditioning | 3,191 | 95 | 3.0% | 171 | 75 |
| Hair and beauty salons | 1,808 | 25 | 1.4% | 0 | 188 |
| Lawyers | 983 | 6 | 0.6% | 0 | 9 |
| Marketing and communication agencies | 1,247 | 5 | 0.4% | 0 | 37 |
| Dentists | 818 | 0 | 0.0% | 10 | 5 |
| Restaurants and bakeries | 1,139 | 0 | 0.0% | 0 | 0 |
| Wedding photographers | 1,040 | 0 | 0.0% | 0 | 12 |
| Total | 18,662 | 1,733 | 9.3% | 360 | 498 |
Recruiters are brands, dentists are people
The share of networks follows the trade more than the country. Recruitment: 709 of 1,429 recommendations go to Hays, Robert Half, Michael Page, Randstad and their peers, in all four countries. Accountants: 22.4%, mostly the large audit firms, plus online accountants (Contabilizei in Brazil, Dougs in France). Insurance brokers 19.4%, estate agents 14.8%, driving schools 9.3%, garages 6.5%. At the other end are the trades where the customer meets a person: lawyers 0.6%, hair and beauty 1.4%, and none at all for dentists, restaurants and bakeries, or wedding photographers. By country, France has the most networks and Brazil the fewest. By engine, ChatGPT names a network in 10.8% of its recommendations, Gemini in 7.5%.
| Country | Recommendations | National networks | Share |
|---|---|---|---|
| United States | 4,928 | 498 | 10.1% |
| United Kingdom | 4,748 | 420 | 8.8% |
| France | 4,356 | 634 | 14.6% |
| Brazil | 4,630 | 181 | 3.9% |
| Total | 18,662 | 1,733 | 9.3% |
| Trades | Networks named |
|---|---|
| Recruitment and staffing agencies | Hays, Robert Half, Michael Page, Randstad, Expectra |
| Accountants | BDO, KPMG, Deloitte, Grant Thornton, In Extenso, Contabilizei |
| Insurance brokers | Aon, Marsh, HUB International, Verspieren, Seguralta |
| Estate and letting agents | Orpi, Century 21, Guy Hoquet, Lopes, Savills |
| Driving schools | Red Driving School, AA, BSM, ECF, Ornikar |
| Garages and auto repair | Midas, Norauto, AAMCO, Speedy, Kwik Fit |
| Plumbers, electricians, locksmiths, heating, air conditioning | Roto-Rooter, Benjamin Franklin Plumbing, Mr. Electric, Pop-A-Lock, Keytek, Cham |
Names that look local, and are not quite
Two other patterns show up among the names given in five cities or more. 360 recommendations are a trade and a city glued together: "Houston Driving School", "Austin Driving School", "Locksmith Leeds", "Plombier Paris", "Chaveiro 24 Horas Campinas". Some businesses with such names exist; what we measured is the same shape returning city after city. They cluster in the trades that come to you (171), driving schools (77) and garages (69), and 299 of the 360 come from ChatGPT. 498 recommendations are one name given in five cities or more without being a network we know: "Garage de la Gare" in nine French cities, "The Cutting Room" in ten British ones, the same dentist's name in five French cities, as our October 7 edition found. Some are common names shared by unrelated businesses; some are names the AI carries from one city to the next. Hair and beauty has the most (188).
What it means if you are the business
If you are independent in a trade where the AI answers with brands (recruitment, accounting, insurance, estate agency), part of the list you compete with is national, and the same brands come back city after city. What sets you apart has to be written just as plainly: the city and neighbourhood you serve, your specialty, the people who do the work. If your trade is one where the AI almost never names a network (dentists, restaurants, photographers, salons, lawyers), your competition in the answer is other independents of your city. And if your business name is only your trade and your city, the AI may give a name of the same shape that is not yours: your full name, address and site, written the same way everywhere, are what tell them apart. Guarantees, as we saw, also mostly belong to networks (AI cites a guarantee: two in three belong to national networks). Nothing here guarantees being named.
What this does not prove
- These are the models' answers through their APIs, without web search: the way our audit asks. Consumer apps with search switched on may answer differently.
- A network here is a name read by hand among those given in at least five of a country's 13 cities. Names seen in three or four cities (2,181 recommendations) were not read: every network share is a floor.
- "Same name, not a known network" mixes common names shared by unrelated businesses and names the AI repeats; we did not check each business.
- We did not check whether each trade-and-city name exists; we counted the shape.
- Four studies collected between July 30 and September 13, 2026, one pass per question. The models have moved since.
Frequently asked questions
Does AI recommend chains or independent businesses?
Which networks does AI name most often?
Why does AI give names like "Houston Driving School"?
Where do these figures come from?
Today's article, by email
At most one email a day. One-click unsubscribe. Your address is used for this and nothing else. Privacy