Tell ChatGPT your furnace stopped working, ask which company to call, and it names about three businesses, against about five on our other HVAC questions. Burst pipes went the same way in our Cheers Market Baseline panel from Oct. 9 to 31, 2026. For an after-hours shop, that's a shorter list to make, and on urgent questions Google's engines print a phone number far more often, so it had better ring someone who can roll a truck.
Last verified: November 1, 2026.
Use these October results to prepare for the heating season. Run the six checks in the cold-snap section below for each branch, then repeat your own visibility checks through the winter.
Two emergency prompts produced shorter lists on average
Two of our six urgent questions sound like a homeowner whose heat just quit or whose pipe just burst: "My furnace stopped working, which local company should I call?" and a burst or frozen pipe that needs a plumber "right now." Across 60 US cities from Oct. 9 to 31, ChatGPT named 3.22 businesses per answer on the furnace question against 4.97 on other HVAC questions, 35% fewer, and 3.02 on the pipe against 5.08 on other plumbing questions, 41% fewer. The median was 3 on both, against 5 for each trade.
The other four urgent questions ask for the best emergency or same-day providers, and their lists barely moved. Emergency electricians available today got 4.72 names against 4.86 for other electrical questions and same-day AC repair got 4.96, a hair under the HVAC average, while emergency plumbing service and same-day plumbing repair came in slightly above plumbing's at 5.26 and 5.29.
Cheers Market Baseline
Describe the emergency and ChatGPT names three companies, not five
Cheers Market Baseline, ChatGPT answers in 60 US cities, Oct. 2 to 31, 2026; queried Nov. 1, 2026
| Urgent question | Position | Businesses named per answer |
|---|---|---|
| My furnace stopped working | Other HVAC questions | 4.97 |
| My furnace stopped working | This question | 3.22 |
| Burst or frozen pipe, right now | Other plumbing questions | 5.08 |
| Burst or frozen pipe, right now | This question | 3.02 |
| Emergency electricians available today | Other electrical questions | 4.86 |
| Emergency electricians available today | This question | 4.72 |
| Same-day AC repair | Other HVAC questions | 4.97 |
| Same-day AC repair | This question | 4.96 |
| Emergency plumbing service | Other plumbing questions | 5.08 |
| Emergency plumbing service | This question | 5.26 |
| Same-day plumbing repair | Other plumbing questions | 5.08 |
| Same-day plumbing repair | This question | 5.29 |
That split is the part an emergency shop should care about. Showing up when someone asks for the best emergency or same-day providers in town doesn't mean you'll show up when they type that the heat quit and ask whom to call.
Both short-list questions describe a breakdown and ask whom to call. Include that wording in your checks: it is a useful test of whether the branch appears when a homeowner needs a decision immediately. The method keeps those two prompts separate from the broader emergency category.
The gap is older than ChatGPT's Oct. 9 switch to a new answer layout. From Oct. 2 to 8, the furnace question got 4.10 names per answer against 5.50 for other HVAC questions and the burst pipe got 3.46 against 5.60, and the gap came back in every week with enough readable answers to count.
Google AI Mode's business cards, which don't depend on ChatGPT's list at all, show the same pattern: 3.58 cards per answer on the furnace question against 4.31 on other HVAC questions, and 3.17 on burst pipes against 4.15 (US, Oct. 9 to 31).
Pool all six urgent questions and the shrink is small. Each engine shows about 0.3 fewer business cards on urgent answers than on other plumbing, HVAC and electrical questions (ChatGPT 4.13 against 4.48), and most of that gap comes from the two crisis questions.
On urgent questions, Google's AI prints a phone number far more often
Gemini put a phone number in 44.8% of urgent US answers in October against 5.1% on planned-job questions, and Google AI Mode in 32.6% against 5.4%. The gap held inside every trade and was widest for electricians, where Gemini printed a number in 76.0% of urgent answers against 10.5% for other electrical questions (US and Canada, Oct. 9 to 31).
Phone numbers in the answer
On urgent questions, Google's AI prints a phone number
- Gemini: a phone number in 5.1% of US answers to planned-job questions, 44.8% on urgent ones
- Google AI Mode: 5.4% on planned-job questions, 32.6% on urgent ones
Cheers Market Baseline, 60 US cities, Oct. 2 to 31, 2026 (about 1,620 urgent and 5,040 planned answers per engine); the US and Canada figures in the caption add 40 Canadian cities. Computed Nov. 1, 2026
| Engine | Planned jobs | Urgent |
|---|---|---|
| Gemini | 5.1% | 44.8% |
| Google AI Mode | 5.4% | 32.6% |
We counted any phone-shaped number in the answer, and on Gemini and AI Mode we didn't check whose it was, so some furnace and burst-pipe answers could be carrying a gas or water utility line. Electrician answers, where a utility number is least likely, show the widest gap.
ChatGPT used to behave like Google's engines, printing a number in 58.5% of urgent answers and 7.7% of the rest from Oct. 2 to 8 (US and Canada). Sistrix found it switched to a building-block layout overnight on Oct. 8 to 9, 2026, and used it in 97% to 100% of sampled answers from Oct. 23. After the switch, 92.5% of US ChatGPT answers in our panel carried at least one phone number (Oct. 9 to 31). Its October average blends the two layouts, so the chart above leaves ChatGPT out.
Whose number is it? We checked that for ChatGPT in a separate study: on urgent questions, 12.4% of the checkable numbers weren't on the named business's own website in the pages we checked, about the same as on planned questions (11.1%; US and Canada, Oct. 2 to 31).
One thing didn't move. Google AI Mode linked a booking or scheduling page in 33.6% of urgent US answers and 33.9% of answers to other questions in the same three trades (Oct. 9 to 31), so on urgent questions what changes is the phone number.
Urgent answers mention 24/7 more and warranty and price less
Urgent answers mention 24/7 or after-hours service almost every time: 99.0% of ChatGPT's urgent US answers in October, 96.6% of Gemini's and 91.7% of AI Mode's, against 36.8% of ChatGPT's planned-job answers. It was about as high in September, and the HVAC guide has that month's split question by question.
Warranty talk falls off. September can't be lined up against October (Method explains why), so the fair comparison stays inside October and inside each trade. From Oct. 9 to 31 (US and Canada), ChatGPT mentioned a warranty in:
- 35.1% of urgent HVAC answers, against 81.5% for other HVAC questions
- 44.9% of urgent plumbing answers, against 90.9% for other plumbing questions
- 60.0% of answers to the emergency electrician question, against 83.3% for other electrical questions
Across all nine engine-and-trade pairs (US and Canada, Oct. 9 to 31), urgent questions got 28% to 61% fewer warranty mentions, and they ran lower in every trade before Oct. 9 as well.
What the answer talks about
On urgent questions, warranty and price come up far less often
ChatGPT
Said more often on urgent questions
- 24/7 or after-hours service: 99.0% of ChatGPT's urgent answers, against 36.8% for planned jobs (US, October)
- A license, insurance or bonding, emergency electricians: 91.3% of ChatGPT's answers, against 46.5% for other electrical questions (US and Canada, Oct. 9 to 31)
ChatGPT
Said less often
- A warranty, HVAC: 35.1% of ChatGPT's urgent answers, against 81.5% for other HVAC questions (US and Canada, Oct. 9 to 31)
- A warranty, plumbing: 44.9%, against 90.9% for other plumbing questions (US and Canada, Oct. 9 to 31)
- A dollar figure: 7.6% of ChatGPT's urgent answers, against 24.8% for planned jobs (US, October)
Cheers Market Baseline, 60 US cities, October 2026; trade lines 100 US and Canadian cities, Oct. 9 to 31, 2026; September line from earlier Cheers articles, Aug. 31 to Sep. 20, 2026; computed Nov. 1, 2026
Price drops away too. ChatGPT put a dollar figure in 7.6% of urgent answers against 24.8% on planned jobs, and financing came up in 7.0% of Gemini's urgent answers against 14.0% (US, October).
Licensing goes the other way on ChatGPT, where 36.2% of urgent answers mentioned a license, insurance or bonding against 22.4% on planned jobs (US, October). On the emergency electrician question it reached 91.3%, against 46.5% for other electrical questions (US and Canada, Oct. 9 to 31), while Google AI Mode's urgent answers mentioned licensing no more often than its planned-job answers (12.3% against 13.9%, US, October). The electrician guide breaks down how emergency and planned electrical work pull apart on permits and code.
These are pattern counts, so "call a licensed plumber" counts the same as a line about a specific company, and a license mention doesn't mean the engine checked anyone's license.
What to check before the first cold snap
EIA's Winter Fuels Outlook, published Oct. 6, 2026, runs its household heating tables from November through March, and natural gas was the main heating fuel in 46% of US homes in 2024. The furnace calls are close. Before they start:
- In your busiest market (your three busiest, if you run several), ask ChatGPT, Gemini and Google AI Mode the crisis version for your trade ("My furnace stopped working. Which local company should I call in [city]?" or "I have a burst pipe. Which plumber should I call right now in [city]?") and the "best emergency" version, for example "Who are the best emergency HVAC companies in [city]?" Run each one three times on different days and mark whether the branch is named each time. A branch that makes the five-name list can still miss the three-name one.
- Write down the phone number each engine prints next to the branch. If it's one of yours, call it after 8 p.m. and note who picks up. If it isn't yours, don't dial it; find where that number is listed and fix that listing first.
- Make every number tied to a branch ring someone who can dispatch at night: the Business Profile's primary number, the location page, each directory listing and every call-tracking number. The call-tracking guide covers what to do when the number an engine prints is a tracking line.
- Brief your CSRs on Gemini Call for Me, using the caller rule below.
- Leave the Business Profile setting "Bookings and inquiries from customers" on wherever a branch can answer booking and price questions. It's a separate switch from Call for Me.
- If you run many locations or report branch visibility to an operating partner, keep crisis-phrased, "best emergency" and planned questions as three separate rows per branch, and re-run all three in December.
Nearly every urgent answer already says 24/7, so the claim alone won't set a branch apart; what matters is that it's true at each branch and that Google can read it. Google's "24 hours" setting is chosen day by day, so a branch with an evening cutoff should show the cutoff. On Sept. 24, 2026, Search Engine Roundtable reported a Business Profile test called "Set your hours automatically," where the owner describes the hours in plain text and Google extracts them, so if that box shows up on your profile, read back what Google pulled out before you leave the page.
In the profile itself, state availability as a fact. Google's Oct. 6, 2026 "For merchants" clause bars "direct calls-to-action" in the listing (the date is Search Engine Roundtable's first report, since Google's page has none). Where to publish emergency hours shows where the after-hours promise belongs on your profile and location pages. Answering at 2 a.m. and arriving at 2 a.m. are different promises, and the response-time standard covers how to publish the second one.
Google is also testing an assistant that phones businesses for the customer. TechCrunch reported Google's test of Gemini Call for Me on Sept. 24, 2026, and Google's help page limits it to US adults with a Pixel 11, a Google AI subscription and the Phone by Google public beta, saying Gemini can ask a business for hours, availability and "service quotes" and book appointments. Google says Gemini starts every call by disclosing that it's an AI assistant from Google calling on a recorded line, and the call comes from the customer's own number.
That gives CSRs a simple caller rule. Treat a Call for Me call as that customer's inquiry and give the hours, availability and price range you'd give any caller. Google also says Gemini can't make payments or relay sensitive personal information, so a "Google" caller who asks for card details, a verification code, a login or a Business Profile change gets treated like any other suspicious caller.
The Business Profile setting works differently: "Bookings and inquiries from customers" covers calls Google places when customers use Search or Maps. Google says its automated calls for bookings, wait times and price checks aren't available in Indiana, Louisiana, Minnesota, Montana or Nebraska (Business Profile Help, read Nov. 1, 2026), and it hasn't said whether the setting applies to Gemini's calls.
Garage door and restoration companies won't find their trades here, because neither one is among our six urgent questions. Google's Local Services Ads, a separate product from these AI answers, do split the two: their screening rules put garage door work in "an urgent category" with added background checks, and water damage isn't on that list.
Make the promise and the response match. When an answer puts your number in front of a homeowner with no heat, the branch needs to answer and give them a workable next step.
Method
The Cheers Market Baseline is our own research panel: every week it asks ChatGPT, Google AI Mode and Gemini the same 32 home-service questions in 100 metro areas, 60 in the US and 40 in Canada. This study uses the 60 US cities from Oct. 2 to 31, 2026 (8,217 answers per engine). The six urgent questions are a burst or frozen pipe, a furnace that stopped working, emergency electricians available today, emergency plumbing, same-day AC repair and same-day plumbing repair. We compare their answers (1,620 per engine) with other questions about the same trade (4,917) and with 20 planned-job questions across all trades (about 5,040); each question runs on its own weekday, so from Oct. 9 to 31 each got three or four runs per city.
From Oct. 9 to 31 we couldn't read ChatGPT's list of local businesses on 31% of furnace answers, 26% of burst-pipe answers and 16% of planned answers and left those out rather than count them as zero, so the furnace average rests on 166 of 240 answers (162 from the Oct. 16, 23 and 30 runs) and the burst-pipe average on 133 of 180. In the week before Oct. 9 we could read 60 of 60 furnace lists and 59 of 60 burst-pipe lists.
Cards are the business cards shown with each answer, averaged over answers that had them: ChatGPT 1,311 urgent and 4,017 other answers, AI Mode 1,468 and 4,527, Gemini 737 and 1,961. We counted Gemini cards from Oct. 2 to 25 only, since no later Gemini answer in our panel carried place cards; we haven't confirmed whether Gemini changed its format or our capture did. Pooled over the six urgent questions, AI Mode showed 3.88 cards against 4.18 and Gemini 3.30 against 3.63, and urgent came out lower in all nine engine-and-trade pairs, by 0.12 to 0.52 cards, though from Oct. 26 to 31, when the burst-pipe question didn't run, the ChatGPT and AI Mode gap shrank to under a tenth of a card.
Phone numbers, 24/7, licensing, warranty, price and financing are text pattern counts: any phone-shaped number, and any mention, general advice included. We checked whose numbers ChatGPT printed from Oct. 2 to 31, not Gemini's or AI Mode's: 12.4% of 314 checkable urgent-question numbers and 11.1% of 334 planned ones weren't on the business's own site, and across question types the US share was 17.6% (66 of 374) against 5.9% in Canada. Gemini's share of all US answers with a number rose from 7.5% in October's first, short week to 21.6% in the last as place cards dropped out of what we captured (since each question runs on its own weekday, short weeks mix questions differently), but its urgent gap came first, at 24.0% against 0.8% of all other answers from Oct. 2 to 8 (US and Canada). Over the whole of October, before and after its change, ChatGPT put a number in 84.9% of urgent answers against 71.5% on planned-job questions; after the switch, 5,835 of 6,307 US answers carried one.
Most figures are US-only; the Oct. 2 to 8 phone shares, the shares inside each trade after Oct. 9 and the whose-number shares pool the US and Canada. ChatGPT's per-question counts cover Oct. 9 to 31, after its layout switch, and its warranty mentions across all US answers went from 48.4% the week before to 79.4% after. The September figures come from earlier Cheers articles (four emergency questions, Aug. 31 to Sep. 20, US and Canada), so ChatGPT's 3.6% warranty share then reflects a different layout and question group. October is before peak heating season, the two crisis questions can't separate urgency from phrasing, we have no call, booking or revenue data, and Perplexity isn't in this edition.
Cite as: Cheers Market Baseline, urgent vs planned home-service questions, 60 US cities, October 2026.
Sources
- What changed in ChatGPT in October: GPT-6, UTM tracking and sources, SISTRIX, October 30, 2026. ChatGPT's Oct. 8 to 9 layout switch.
- Winter Fuels Outlook, U.S. Energy Information Administration, October 6, 2026. Heating season months and natural gas share of homes.
- Edit your business hours, Google Business Profile Help, undated. Day-by-day "24 hours" setting.
- Google Business Profiles Tests Set Your Hours Automatically, Search Engine Roundtable, September 24, 2026. The automatic hours test.
- Maps user-contributed content policy, "For merchants", Google, undated. The ban on direct calls-to-action in listings.
- Updated Google Maps Policy: What Merchants Can't Post In Listings Or Posts, Search Engine Roundtable, October 6, 2026. First-report date of the merchant clause.
- Google tests letting Gemini call businesses for you, TechCrunch, September 24, 2026. Call for Me test report.
- Ask Gemini to handle your everyday phone calls, Google Gemini Help, undated. Call for Me eligibility, disclosure and limits.
- About automated calls and texts from Google to your business, Google Business Profile Help, undated. Automated business calls and excluded states.
- Business screening and verification requirements, United States, Google Local Services Help, undated. Urgent-category screening for Local Services Ads.
Sources read November 1, 2026.
Dylan Allen-Arnegård is the CEO & Co-Founder of Cheers, the done-for-you local visibility partner for established home-service companies and PE-backed portfolios.