Matt Turley, Continuum. Updated October 2, 2026.
One city, one day, 10 ChatGPT answers, every company and source logged. The three-run Perplexity and Gemini numbers from the same 10 questions sit beside it so you can see how the tools differ.
Yes, by name. Asked 10 questions a Dallas homeowner asks about AC and heating, once each on October 1, 2026, ChatGPT named 42 different HVAC companies, and Rescue Air and Plumbing was named in 9 of the 10 answers. Which companies it names depends on which sites it reads first: the sites it cited most were dfwprofessionals.com (6 of 10 answers), expertise.com (6 of 10 answers) and consumeraffairs.com (5 of 10 answers).
These are the companies written in the answer text, counted once per answer. A link or a citation alone does not count. The questions were the plain things a homeowner types: best HVAC company in Dallas, who to call for emergency AC repair, an honest AC company that will not upsell, a family owned one, and so on. The full list of 10 is on the research page.
| Company | Answers naming it (of 10) | Share |
|---|---|---|
| Rescue Air and Plumbing | 9 | 90.0% |
| Texas AirZone LLC | 8 | 80.0% |
| Berkeys Plumbing, A/C & Electrical | 6 | 60.0% |
| Cody & Sons Plumbing, Heating & Air | 6 | 60.0% |
| Quality 1 Energy Systems Heating & Air Conditioning | 6 | 60.0% |
| Bacon Plumbing Heating Air Electric | 5 | 50.0% |
| Coolinary Arts Heating & Cooling | 5 | 50.0% |
| Willard Cooling, Heating, Plumbing, & Electrical | 5 | 50.0% |
| Air Patrol | 4 | 40.0% |
| Cozy - D Heating & A/C LLC | 4 | 40.0% |
| Quigley Heating & Air Conditioning | 4 | 40.0% |
| Serveway Heating & Air Conditioning | 4 | 40.0% |
| Total Air & Heat Company | 4 | 40.0% |
| Ellis AC & Furnace Repair | 3 | 30.0% |
| On Time Experts | 3 | 30.0% |
Not a ranking of anyone's work. These are public companies in a public answer, logged as given. The other 27 companies were named in fewer answers and are in the JSON file.
ChatGPT searches the web, reads a handful of pages, and names the companies written up on them. Across the 10 answers it cited 66 sources, about 6.6 per answer. 50.0% of those citations went to company websites and 36.4% to directories and review sites. The sites below are the ones to be listed on, with your name spelled the same way everywhere.
| Site | Type | Answers citing it (of 10) |
|---|---|---|
| dfwprofessionals.com | Directories and review sites | 6 |
| expertise.com | Directories and review sites | 6 |
| consumeraffairs.com | Directories and review sites | 5 |
| dallasacrepairguide.com | Company websites | 5 |
| reddit.com | Reddit, forums, social and video | 4 |
| consumerbusinessreview.org | Company websites | 3 |
| homeadvisor.com | Directories and review sites | 3 |
| airco.com | Company websites | 2 |
| dallastoprated.com | Directories and review sites | 2 |
| dirttrades.com | Company websites | 2 |
Source types, counted once per answer: company websites 50.0%, directories and review sites 36.4%, best-of lists and comparisons 7.6% and reddit, forums, social and video 6.1%.
The same 10 questions went to Perplexity and Gemini three times each on October 1, 2026, 60 answers in all. There a company only counts on a question when it is named in at least 2 of the 3 runs, which is a stricter test than the single ChatGPT run, so the two sets of numbers are shown side by side and never added together.
| AI tool | Answers | Companies named | Named most | Citations per answer |
|---|---|---|---|---|
| Perplexity (sonar), 3 runs, majority rule | 30 | 28 | Berkeys Plumbing, A/C & Electrical, 6 of 10 questions | 18.0 |
| Gemini (gemini-2.5-flash), 3 runs, majority rule | 30 | 35 | Rescue Air and Plumbing, 9 of 10 questions | 7.2 |
| ChatGPT (gpt-4o), 1 run, counted on its own | 10 | 42 | Rescue Air and Plumbing, 9 of 10 answers | 6.6 |
Perplexity and Gemini agreed on only 12 of the 51 companies they named at least once. Their most-cited sites were bestpickreports.com (48 of 60 answers), consumeraffairs.com (38 of 60 answers) and reddit.com (31 of 60 answers). Of the companies ChatGPT named, 5 also made the three-run majority list for the other two tools, which is a hint that the sites feeding all three overlap. Each tool reads a different mix of sites, so being named on one says little about the others. The full three-run tables are on the research page, and the whole series is at Who AI recommends in 2026.
ChatGPT: the OpenAI API, model gpt-4o, with the web search tool forced on, asked each of the 10 questions once on October 1, 2026 (run r-20261001T2351Z-rv26-hvac-dallas-2b93). Cited URLs come from the answer annotations. Perplexity and Gemini: their APIs, each question three times on October 1, 2026, a company counted only when named in at least 2 of the 3 runs. Nobody was signed in and there was no system prompt: the question was the whole input.
Limits, stated plainly: one city, one day, 10 ChatGPT answers. One run per question is a thin sample and the names move between runs, so read the ChatGPT table as one look on one day, next to the steadier three-run numbers. Source types come from a written rulebook, not a person reading every page, so a few sites land in the wrong group. A company counts as named when its name is in the answer text with links and citation chips removed; spelling variants (curly apostrophes, hyphens, “&” versus “and”) count as one name.
This page covers one city and one trade. The AI Answer Check runs the same test on your own business: the questions your customers ask, three runs each on ChatGPT, Perplexity and Gemini, whether you are named, who is named instead, which sites they read, and the three fixes worth the most. It is $299, and I read every report before it goes out.
It does not promise what any AI tool will say. It tells you where you stand today, with the answers to back it up.