Skip to main content
Cheers

Research

What Perplexity Searches When You Ask for a Plumber

We logged 56,712 Perplexity searches on home-service questions. None added "near me." From Aug. 6, a few went straight to the BBB and license boards.

Dylan Allen-Arnegard, CEO and Co-Founder of Cheers
Dylan Allen-Arnegård

CEO & Co-Founder, Cheers

15 min readPublished

Last verified .

A blank building permit card in a clear plastic sleeve beaded with rain is stapled to a thick wooden post in front of a house being framed on a wet job site

When our research panel, the Cheers Market Baseline, asks Perplexity for a local contractor, its searches stay close to the question and its follow-ups keep asking about price and licenses. That's the pattern across 56,712 searches it reported between July and October 2026. From August 6, some went straight to bbb.org or a state license board. Your BBB profile and license record are among the pages it looked for, and you can check both this week.

Last verified: November 1, 2026.

Perplexity answered the panel's 32 home-service questions 37,012 times across 100 US and Canadian metros. We could log this engine's searches in full, and its help center says it searches the internet in real time and cites what it finds. The log gives us a direct view of the records an assistant checks while researching a local provider.

Google's guide to generative AI features calls it query fan-out when a system runs several related searches to gather information for one answer. Our explainer on query fan-out in Google AI Mode covers what Google has said; this log shows what one engine typed.

Word-for-word copies of the question made up 43.7% of the log, almost all with the city in, and rewrites that kept the city added 28.4%, so at least seven in ten searches named the city. Web addresses in place of search words came to 4.2% and everything else 4.4%. The remaining 19.3% were bare service terms with no place in them, like "hvac companies" (1,412 times) or "plumbers" (725), and 10,890 of those 10,965 searches landed between July 18 and August 17.

The rewrites stayed on the question's subject. When a question asked about pricing, deals or quotes, 38.0% of its rewrites included a price word, against 0.1% on the other 26 questions, so Perplexity didn't add price on its own. That interest shows up later, in the follow-ups.

"Near me" only shows up when the question says it

None of the panel's questions says "near me", so its count of zero isn't the real test; that needs questions other people wrote.

Twenty companies that use Cheers track 636 local questions that leave the phrase out, and Perplexity ran 1,152,987 searches for them in the same window. It added "near me" to 47, about 4 in 100,000. For the 13 home-service companies alone, it was 38 of 666,522.

When a question did say "near me", 56.0% of Perplexity's searches for it repeated the phrase. The phrase was usually supplied by the question, though the engine also added it in 47 searches.

Perplexity search log

Perplexity almost never added "near me" on its own

47

of 1,152,987 searches added "near me", about 4 in 100,000

Base: 636 questions tracked by 20 companies that use Cheers, across home services and other local categories

56.0%

of searches repeated "near me" when the question used it

191,045 searches, 12 companies; 50.0% without the company with the most of these searches

38

of 666,522 searches added "near me" for home-service companies' questions

13 companies

0 of 56,712

Cheers Market Baseline searches contained "near me"

32 questions, 100 metros; none of the questions says "near me"

Questions that name the tracked company, Cheers-owned accounts and test accounts are excluded. Aggregate only.

Cheers Market Baseline and questions tracked by 20 companies in Cheers, Perplexity search log, Jul 2 to Oct 14, 2026. Pulled Nov 1, 2026.

Perplexity added near me to 47 of 1,152,987 searches for tracked questions that didn't say it, about 4 in 100,000. When the question did say near me, 56.0% of 191,045 searches repeated it, and 50.0% without the company with the most of these searches. For home-service companies alone, 38 of 666,522 searches added it. In the Cheers Market Baseline, 0 of 56,712 searches contained near me.
FindingValueBase
of 1,152,987 searches added "near me", about 4 in 100,00047636 questions tracked by 20 companies that use Cheers, across home services and other local categories
of searches repeated "near me" when the question used it56.0%191,045 searches, 12 companies; 50.0% without the company with the most of these searches
of 666,522 searches added "near me" for home-service companies' questions3813 companies
Cheers Market Baseline searches contained "near me"0 of 56,71232 questions, 100 metros; none of the questions says "near me"

That matters if you or your agency test visibility by typing "furnace repair near me" into Perplexity. Many of its searches will repeat the phrase because your question did, which tells you nothing about how it handles a customer who named the city.

From August 6, Perplexity's searches started naming specific websites

A site: search is limited to one website; type "site:bbb.org" in front of the words and every result comes from bbb.org. None of the 20,160 searches Perplexity reported from July 2 to August 5 used one. From August 6 through October 14, these site checks made up 1,416 of 36,552 searches (3.9%), about one in 26.

Perplexity search log

From Aug 6, some of Perplexity's searches named a single website

Share of logged searches limited to one website

0%6%12%0%Jul 60%Jul 130%Jul 200%Jul 270.9%Aug 30.8%Aug 103.1%Aug 172.4%Aug 242.8%Aug 314.6%Sep 710.6%Sep 142.5%Sep 213.8%Sep 286.8%Oct 5
Weeks start Monday. None of the 20,160 searches before Aug 6 was limited to one website; 1,416 of the 36,552 after were. The partial weeks of Jun 29 (Jul 2 to 5, none) and Oct 12 (106 searches) are left out. The change shows up even when we hold our own collection method fixed, but we can't rule out a change in how Perplexity reports its searches.

Cheers Market Baseline, Perplexity, 56,712 logged searches, 32 questions in 100 US and Canadian metros, Jul 2 to Oct 14, 2026. Pulled Nov 1, 2026.

Share of Perplexity searches limited to one website, by week, Jul 6 to Oct 5, 2026
Week startingShare of searches
Jul 60%
Jul 130%
Jul 200%
Jul 270%
Aug 30.9% (First site checks on Aug 6)
Aug 100.8%
Aug 173.1%
Aug 242.4%
Aug 312.8%
Sep 74.6% (1,531 searches)
Sep 1410.6% (Includes Sep 18 to 20, 43% of all site checks)
Sep 212.5%
Sep 283.8%
Oct 56.8% (1,500 searches)

They didn't arrive evenly. September 18 to 20 alone produced 614 of them (43.4%), inside the 10.6% week, when Perplexity ran about 4.3 searches per answer, so read that peak as a burst. Without those three days, site checks were 2.5% of searches after August 6, and the order of sites barely moved.

The change shows up even when we hold our own collection method fixed, but we can't tell whether Perplexity began running these searches on August 6 or began reporting them.

Promptwatch saw something similar in ChatGPT two days later. Its data, reported by Search Engine Journal on August 19, 2026, put ChatGPT's share of fan-out searches using site: at 0.37% before August 8 and 16.8% from that day, though Promptwatch's own data page calls the figures provisional. That separate dataset adds context to the shift toward named-site searches.

Where the site checks went

On questions about US cities, Perplexity ran 969 site checks. The BBB drew 240 (24.8%), contractors' own websites about 190, and Google and license boards 122 each. Then came Angi with 72, utility and rebate programs about 60, manufacturer installer and dealer pages 55, associations and certifiers 18 and Yelp 16, and the last 73 went elsewhere.

Where the site checks went

The BBB and contractors' own websites drew the most site checks

site: searches on US-city questions

BBB240
Contractors' own websites192

About 190; none searched more than four times

Google122

Many of them Google Maps pages

License boards and government license pages122

California's CSLB led

Angi72
Utility and rebate programs59

About 60

Manufacturer installer and dealer pages55

Led by Tesla and Generac, from the EV charger and generator questions

Associations and certifiers18
Yelp16
Everything else73

Other government pages, other platforms, retailers and unreadable targets

Groups come from published domain lists; contractors' own websites are everything not on a list and are counted, never named. On Canadian-city questions (330 site checks), contractors' own sites (71), Google (70), the BBB (63) and HomeStars (61) led; 117 site checks named no city.

Cheers Market Baseline, Perplexity search log, 32 questions in 60 US metros, Jul 2 to Oct 14, 2026. Pulled Nov 1, 2026.

Perplexity site: searches on US-city questions by kind of site, Jul 2 to Oct 14, 2026
Kind of site searchedsite: searches
BBB240
Contractors' own websites192 (About 190; none searched more than four times)
Google122 (Many of them Google Maps pages)
License boards and government license pages122 (California's CSLB led)
Angi72
Utility and rebate programs59 (About 60)
Manufacturer installer and dealer pages55 (Led by Tesla and Generac, from the EV charger and generator questions)
Associations and certifiers18
Yelp16
Everything else73 (Other government pages, other platforms, retailers and unreadable targets)

Read the contractor group twice. Across the whole log it came to about 280 site checks over 265 websites, none drawing more than four, so a contractor's own website is one of the places Perplexity searched directly. That's one more reason the fixes below put your license number and prices on yours in plain text.

Canada differs in one place: on the 330 site checks for Canadian cities, contractors' own sites (about 70), Google (70), the BBB (63) and HomeStars (61) led, and HomeStars drew none on US questions.

Prioritize these records alongside your own website. After August 6, about 1 search in 120 went to bbb.org, 1 in 200 to a license board and 1 in 650 to a manufacturer page.

Of the 192 google.com checks, 89 pointed at Google Maps pages and 134 included "review" or "rating".

California's Contractors State License Board led the license group with 70 site checks and Texas's TDLR had 12, and 105 of the 180 license-board searches used words like lookup, verify, check or status. The FTC tells homeowners to run that same check before they hire.

Trade changes the mix. On US plumbing questions, the 378 site checks went to the BBB 134 times (35.4%), Google 65, license boards 53 and Angi 45. HVAC leaned on the BBB less, but across all cities, 257 of its 485 site checks came from one question about deals on a new furnace or air conditioner, so read HVAC's mix as that question's mix.

Electrician questions are the outlier. Of their 178 US site checks, license boards drew 42 (23.6%), manufacturer pages 35 and the BBB only 15, which is the practical case for putting license and insurance proof on the page where a buyer, or a search engine, can check it. The manufacturer checks (Tesla 21, Generac 13) came from the EV charger and generator questions.

Citations, a different measure, point a similar way. In a Yelp-commissioned count of organic US AI citations in the fourth quarter of 2025, the BBB came second of six platforms, with 149.7K to Yelp's 512.7K and Angi's 145.6K. Yelp's shareholder letter doesn't name the firm that counted, and what each AI engine cites is a separate question.

What Perplexity suggests asking next

Perplexity suggests follow-up questions alongside its answers, and we sorted the 129,679 we logged into keyword themes that can overlap. Some themes just echo the question; others are topics Perplexity raised on its own.

Urgency is mostly echo: it appeared in 66.3% of follow-ups to the six urgent questions and 4.6% of the rest. Price is different, showing up in 12.6% of follow-ups to the 26 questions that never mention pricing, deals or quotes (29.2% on the six that do). Reviews do something similar, at 27.6% after the four review questions and 10.4% after the rest.

What Perplexity suggested next

Perplexity's follow-ups echo the question, but price shows up anyway

Asked aboutShare of follow-upsNot asked
66.3%4.6%Urgency (emergency, same-day, weekend)29.2%12.6%Price or cost27.6%10.4%Reviews, ratings or complaints
  • Urgency (emergency, same-day, weekend): 6 urgent questions vs 26 others
  • Price or cost: 6 questions about pricing, deals or quotes vs 26 others
  • Reviews, ratings or complaints: 4 review questions vs 28 others
No question mentions a license, yet 13.5% of all follow-ups raised licensing, insurance or permits; 14.2% of follow-ups to the 31 questions that don't ask for quotes compared options or prices. Keyword themes; one follow-up can match several; 1,291 malformed entries removed.

Cheers Market Baseline, 129,679 Perplexity follow-up suggestions across 37,012 answers, Jul 2 to Oct 14, 2026. Pulled Nov 1, 2026.

Perplexity follow-up topics by whether the question raised them, Jul 2 to Oct 14, 2026
TopicAsked aboutNot asked
Urgency (emergency, same-day, weekend)66.3%4.6%
Price or cost29.2%12.6%
Reviews, ratings or complaints27.6%10.4%

Licensing, insurance or permits came up in 13.5% of all follow-ups though no question mentions a license, and in 23.2% on electrician questions. On plumbing questions in both countries, 17.8% of follow-ups asked about price and 11.9% about licensing. Comparisons ran 14.2% on the 31 questions that don't ask for quotes.

These are suggested questions, not proof that a homeowner asked them. Still, the searches stuck to the question while the follow-ups wanted to know what the job costs and whether the contractor holds a license, which is one more reason to publish price ranges.

The pages a branch can write from these follow-ups

Narrower patterns inside those themes map to pages a branch can write. With heating season starting, get the furnace and water heater versions live before the November and December calls; the titles below are examples, not data.

  • Compare quotes, bids or prices (8,709 follow-ups, 6.7%). "How to compare furnace replacement quotes": the lines a fair quote includes and what to ask about each.
  • Cost of a job (7,376, 5.7%). "Water heater replacement cost in Omaha": a range, what moves it, and a price that includes the fees every customer pays.
  • Equipment brands (7,083, 5.5%) and warranty (6,205, 4.8%). "Brands we install and the warranties that come with them", linked to your listing in each manufacturer's locator.
  • "What criteria define the best..." (4,549, 3.5%). "What to look for in an electrician", linked to your full Google reviews, not a hand-picked set.
  • Verify a license or insurance (4,144, 3.2%). "Our license and insurance", with the numbers and links to the board's lookup.
  • Questions to ask before hiring (2,535, 2.0%). "Questions to ask before you hire a roofer."
  • Financing, rebates or tax credits (1,756, 1.4%). A page naming the utility rebate program you work with, since utility and rebate pages drew about 110 site checks. Financing terms go up only with the lender's disclosures.

Don't build from the most repeated strings. The top one, "what criteria define strongest reviews for plumbers in [city]" (106 times), just restates the panel's own question.

Write one page per real question in each market where the answer differs, never one per wording. Google's AI guide warns against separate content for every variation or fan-out query and ties it to its scaled content abuse policy, and its spam policies treat pages "created to rank for specific, similar search queries" as doorways. The rules for city pages without doorway risk show where that line sits.

What to fix this week

  • Check your license record against your name. Look yourself up by business name and license number on the board that regulates you (California's CSLB, Texas's TDLR, or for Ontario electricians the ESA's contractor lookup). If that name doesn't match your website and Business Profile, add your trade name to the board record where allowed, or show the legal name next to the license number on your site; don't rename the Business Profile to a legal name customers don't use. Then print the license number as text on each location page, linked to the board's lookup, as the license and insurance guide lays out.
  • Write the furnace and water heater pages first, from the list above.
  • Put prices and emergency availability in plain text on your website, and only what's true at that branch. A price has to include the fees every customer pays, which California and Minnesota both require. Financing copy that names a payment amount, term or down payment triggers extra disclosures under Regulation Z, the APR among them, and a special needs its terms and an end date. Use the dedicated service-price field for an actual flat fee, as Google's services help permits; keep promotional pricing out of the business description and custom service names.
  • Open the bbb.org profile for each location and correct the name, phone, address, city and service categories. It was the single most searched site in the log. The site-check count differs from the related answer-citation study, so we do not use the two counts interchangeably. Whether accreditation is worth paying for is a separate question.
  • Name the city and the service plainly in your website's titles, headings and first sentence ("Water heater repair in Omaha"), not "near me". Keep them out of your Google Business Profile name, though: Google's guidelines say it should be the real-world name customers know, and unnecessary extra words in it can get a profile suspended.
  • If a branch is a certified installer or authorized dealer (Tesla Wall Connector, Generac, a ductless brand), check that it appears in the manufacturer's locator, such as Generac's dealer locator, with its own phone number and URL. Claim the status only while you hold it and follow the manufacturer's logo rules. Manufacturer pages appeared in 56 site checks, mostly on electrician questions, making the locator particularly relevant for those services.

Method

The Cheers Market Baseline is a research panel Cheers runs, not a set of customers: 32 home-service questions asked every week in 100 metros (60 US, 40 Canada), with the place written into the question. For each Perplexity answer from July 2 to October 14, 2026, we stored the web searches Perplexity reported and the follow-ups it suggested: 37,012 answers, 56,712 searches and 130,970 follow-ups, less 1,291 malformed entries, leaving 129,679. A search counts once for each answer that contained it, and Perplexity reported about 1.5 per answer, a rate that moved between about 1 and 2. On nine days we have searches but no answer record, so per-answer figures are approximate.

The "near me" check uses Perplexity searches for 636 questions across 53 accounts belonging to 20 companies that use Cheers, in home services and other local categories, over the same window. The questions that did say "near me" were 55 questions across 16 accounts belonging to 12 companies, with 191,045 searches; leaving out the company with the most of them drops the 56.0% repeat rate to 50.0%. We count parent companies rather than locations, publish aggregates only, and excluded questions that name the tracked company, Cheers-owned accounts and test accounts.

Weekly site-check shares (weeks start Monday) sat at 0.0% from July 6 through the week of July 27, then ran 0.9% in the week of August 3, 0.8%, 3.1%, 2.4%, 2.8%, 4.6% (September 7, on 1,531 searches), 10.6%, 2.5%, 3.8% and 6.8% (October 5, on 1,500 searches). The chart leaves out the partial weeks at each end: July 2 to 5, with no site checks, and the week of October 12, with 106 searches. September 18 to 20 carry 43% of the site checks; without them, site checks were 802 of 31,816 searches after August 6, led by the BBB (158), contractors' own websites (about 150), license boards (125) and Google (104).

Two other things entered the log that same week: "licensed, insured or certified" in rewrites, none before August 4, and web addresses in place of search words, 2,362 from August 4 against 19 before. Perplexity's changelog has entries dated July 27, August 4 and August 24, 2026, and none mentions a change to how it searches.

Country comes from the city named in a search or follow-up. Themes, site groups and follow-up patterns are keyword and domain matches. Price means cost, price, how much, charge, fee, rate, afford, cheap or budget (charge, fee and rate as whole words); urgency means emergency, 24/7, same-day, after-hours, weekend, tonight, right now or fast; licensing means license, insurance, certified or permit; comparisons means vs, versus, compare, comparison, difference or better; reviews means review, rating or complaint. Another 3,995 follow-ups (3.1%) were in other alphabets, which our English keywords can't read.

Site checks go into groups by domain: the BBB, Google, Angi, HomeStars and Yelp by their own domains; license boards by a list of state and provincial board domains plus government pages searched with a license word; utility and rebate programs, manufacturers, associations and other government pages by their own lists. Contractors' own websites are every target on no list, so that group may hold a few utility pages and its totals carry "about"; we count those sites without naming any.

Limits: this is Perplexity only, and we didn't capture ChatGPT's, Gemini's or Google AI Mode's searches, though Google's AI features page says AI Overviews and AI Mode may use fan-out too. The panel's questions are templated, so real customers word things more loosely, and 117 of the site checks name no city. The log shows what Perplexity searched and suggested, not which pages it read, cited or recommended, and none of it proves a page change would get a business recommended. Themes miss some phrasing, and the last two weeks are thin.

Sources

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.

Frequently Asked Questions

About 1.5 on average in our log of 37,012 answers to home-service questions, July 2 to October 14, 2026, though the rate moved between about 1 and 2 over those months and spiked near 4 for three days in September. Word-for-word copies of the question made up 43.7% of the searches.

Almost never. Across 1,152,987 searches for questions tracked by 20 companies that didn't say "near me", Perplexity added it 47 times, about 4 in 100,000. When the question did say it, 56.0% of searches repeated it. At least seven in ten panel searches named the city, so put your city and service in your website's titles and headings.

These are records where a buyer can verify a business and its credentials. On US-city questions, the BBB drew 240 of 969 searches limited to one site, and license boards drew 122. Those checks began appearing in our log on August 6, 2026. Keep the branch name, license status and contact details accurate on each record.

Write one useful page for each real buyer question in each market where the answer differs, such as what a job costs or how to check your license. Google's AI guide warns that separate content for every variation or fan-out query, made mainly to manipulate results, breaks its scaled content abuse policy. Use the follow-up counts to pick topics, not as a list of pages to generate.

Yes. Accurate license records, directory profiles and useful service pages support the whole buying journey. This study follows Perplexity specifically; the linked ChatGPT and Google research covers those engines. Use each engine's actual answers to decide which records need attention in your market.

Written by

Dylan Allen-Arnegard, CEO and Co-Founder of Cheers

Dylan Allen-Arnegård

CEO & Co-Founder, Cheers

Dylan leads Cheers, the done-for-you local visibility partner for established home-service companies and PE-backed portfolios. He started out helping frontline teams earn reviews.

Share this article

We logged 56,712 Perplexity searches on home-service questions. None added "near me." From Aug. 6, a few went straight to the BBB and license boards.

Next step

Is AI recommending your business?

Find out how visible you are across ChatGPT, Gemini, Perplexity, and AI Overviews.

Keep reading