Skip to content
Waseit
← All articles
AEOGEO

In the US, your trade decides who AI names

872 answers from ChatGPT and Gemini across 13 US cities. For staffing agencies, 74% of citations go to names that come up in three cities or more. For attorneys, dentists and HVAC contractors: 0%. What that changes for a local business.

74%

of staffing agency citations go to names found in 3 cities or more

September 14, 20268 min read
In this article
  1. What we measured
  2. Fifteen trades, three landscapes
  3. Where national brands hold the answer
  4. Where no name travels
  5. Attorneys and dentists: the AI names less often
  6. A name in several cities is not always a chain
  7. What this measurement does not prove
  8. What it changes for you

Short answer: in the US, "who will the AI name?" has a different answer depending on the trade. For a staffing agency, 74% of business citations go to names that come up in at least three of the thirteen cities we studied: Robert Half is cited in all thirteen. For an attorney, a dentist or an HVAC contractor, that figure drops to 0%: every name cited is local, and it changes from one city to the next. These are not two versions of the same problem. They are two different problems.

0%

for attorneys, dentists and HVAC contractors

67%

of answers about dentists name a business at all

In some trades the AI answers with the same brands everywhere. In others it never knows anyone twice. You do not approach the two the same way.

What we measured

The data comes from the Waseit US study published on September 13, 2026: 872 answers from ChatGPT and Gemini to the two buying questions our audit asks ("what are the best X in Y?", "list the most recommended X in Y"), for 18 trades in 13 US cities. Each trade-city pair got four answers: two questions, two engines.

For this article we counted every business cited and checked in how many cities its name comes back for the same trade. A name present in three cities or more is what we call a name that travels. Directories (Avvo, Zocdoc, Zillow...) are counted separately, and the advice headings that US answers slip into their lists ("Communication Skills", "Referrals") are left out. The table covers the 15 trades measured in all 13 cities. Hotels are excluded because their names carry the city ("Four Seasons Hotel Seattle"), so one chain would count as thirteen different names.

Fifteen trades, three landscapes

Waseit US study, collected September 13, 2026: 52 answers per trade (13 cities, 2 questions, ChatGPT and Gemini). "Names that travel": share of business citations going to a name cited in at least 3 cities for that trade. "Agreement": average share of names both engines have in common in the same city.
TradeAnswers naming a businessDifferent businesses citedNames that travelChatGPT-Gemini agreement
Staffing agency96%9274%11%
Accountant96%14738%6%
Insurance broker98%19338%5%
Driving school90%15337%2%
Locksmith87%13123%1%
Hair salon79%17623%1%
Plumber92%17020%0%
Auto repair shop98%23619%1%
Real estate agent77%17710%2%
Electrician77%1758%1%
Wedding photographer98%2347%1%
Marketing agency98%2492%4%
Attorney73%1860%0%
Dentist67%1370%1%
HVAC contractor85%1840%1%

Where national brands hold the answer

Three trades stand out. Staffing agencies first: Robert Half is cited 37 times and in all 13 cities, Aerotek 27 times in all 13, Insight Global 26 times in 12. Only 92 different businesses for the whole country, the lowest figure in the table. Then accountants (RSM US cited in 9 cities, Deloitte in 7, PwC in 5) and insurance brokers (HUB International in 10 cities, Aon in 6): in both, 38% of citations go to names that travel.

The trades are not spared. Among plumbers, two networks come out: Roto-Rooter, cited in 10 cities, and Benjamin Franklin Plumbing, in 8. Among auto repair shops, Midas and AAMCO. That is not most of the citations (20% and 19%), but it is a seat taken in a list of a few names, city after city.

This is also where the two engines agree most: 11% of names in common for staffing agencies, against 0% for plumbers. And of the 27 names ChatGPT and Gemini share in the same city for that trade, 26 are names that travel.

Where no name travels

At the other end: attorneys, dentists and HVAC contractors, 0%. Not a single name cited in three cities. Marketing agencies are close behind at 2%, with 249 different businesses in 52 answers, the most scattered landscape in the study.

In these trades, no national brand holds a seat. But there is no local consensus either: for attorneys, ChatGPT and Gemini have on average no name in common in the same city. The list is open, and it moves from one engine to the other.

Attorneys and dentists: the AI names less often

Two regulated trades show another sign. Only 67% of answers about dentists name a business, and 73% for attorneys, the two lowest rates in the table. The rest of the time, the assistant answers with a list of criteria to check rather than with names. Directories do appear, but rarely: Avvo, Martindale-Hubbell, FindLaw, Super Lawyers and Chambers and Partners add up to 7 citations for attorneys, and Zocdoc comes up for dentists.

For a practice, the consequence is concrete: one answer in three or four names nobody, and the ones that do name someone cite local practices, different in every city. No brand holds the seat.

A name in several cities is not always a chain

The "names that travel" column needs one precaution. "A1 Driving School" comes back in 12 cities, "A-1 Locksmith" in 11. These are not necessarily networks: plenty of independent businesses carry those names, with no link between them. For driving schools and locksmiths, the figure therefore also measures namesakes. For staffing agencies, accounting firms and brokers, the names at the top are indeed national networks.

What this measurement does not prove

That a brand will always be cited.

Four answers per city and trade is a snapshot, not a stable average.

What users see.

The answers come from the engines' APIs, not from the app a customer opens, which can add a web search.

Every assistant.

ChatGPT and Gemini only. Perplexity was not asked for this study.

An exact count.

Names are extracted by our parser, then cleaned; we would rather miss a few than count an advice heading as a business. The figures slightly undercount.

What it changes for you

If your trade is held by networks

(staffing, accounting, insurance): the goal is not to replace Robert Half or HUB in the answer, it is to appear next to them, on the precise query for your city and your specialty. Ask it yourself, on both engines.

If no name travels

(attorneys, dentists, HVAC, agencies): nobody holds the seat for long, and the two engines do not agree. It is the most open ground, provided you are readable on both.

If you are an attorney or a dentist

: the assistant is more reluctant to name anyone. When it does, it names a local practice; be the one whose city, specialty and reviews it can read, and check your profile on the directories it cites (Avvo, Zocdoc).

All of this can be checked without us, by asking ChatGPT and Gemini your customers' question. Our free audit does the same across several questions and engines, and lists the names cited instead of you.

Frequently asked questions

Which trades are dominated by national brands in ChatGPT and Gemini answers in the US?
In our September 2026 study, staffing agencies first: 74% of citations go to names found in at least three of the 13 cities, with Robert Half and Aerotek cited in all 13. Accountants and insurance brokers follow at 38%.
Can a local business be cited by AI against the chains?
Yes, and its odds depend on the trade. For attorneys, dentists and HVAC contractors, no name is cited in three cities or more: every citation goes to a local business. The two engines agree very little, though, so you need to be visible on each.
Why does AI name an attorney or a dentist less often?
In regulated trades, assistants name a business less often: 67% of answers about dentists and 73% about attorneys name one, against 96% to 98% in other trades. The rest of the time they answer with criteria to check. Directories such as Avvo, Martindale-Hubbell, FindLaw or Zocdoc also appear, but rarely.
How many answers are these figures based on?
872 answers from ChatGPT and Gemini collected on September 13, 2026, for 18 trades in 13 US cities, which is 52 answers per trade. It is a sample, not a census, and the method is public.