The State of Plumbing Search in Florida (2026)

The finding in one sentence

Ask three AI assistants for the best plumber in a Florida city and you will get three different answers in most of them: in zero of the 16 cities we measured in July 2026 did Grok, Gemini, and Claude name the same top plumber, and the plumber with the most Google reviews showed up in just 21% of AI answers on average.

What we measured

This is the Q3 2026 plumbing edition of the Florida Local Search Index, collected on July 29, 2026. We asked each engine a frozen set of 8 plumbing questions, 5 times each, across the same 16 Florida cities as the roofing edition, and cross-referenced every answer against the live Google Places review leaders in each city (309 top-20 listings, 292 unique businesses).

One honest disclosure before the numbers. We planned four engines. Grok, Gemini, and Claude completed all 640 answers each (1,920 recorded answers). OpenAI captured only 254 of its 640 planned answers before that account's prepaid credit ran out mid-run, so its panel is partial. Every headline cross-engine statistic in this report uses the three complete engines. The two layers that pool everything we recorded (citations and winner-site signals) are computed across all 2,174 recorded answers and labeled that way in the sentence. Everything below is measured, recorded, and reproducible. Nothing is an estimate.

Who AI recommends: the review leader is a coin you lose 4 times out of 5

Across the 16 cities we measured in July 2026, the plumber with the most Google reviews appeared in 21% of AI answer runs on average across the three complete engines. In 6 of the 16 cities the review leader appeared in fewer than 10% of runs. In one city it never appeared at all.

The spread is the story. Review leaders in Fort Myers (54%), Lakeland (43%), and Sarasota (41%) get real AI visibility. Review leaders in Cocoa (1%), Hollywood (3%), Tampa (6%), Cape Coral (7%), and Gainesville (7%) are close to invisible, and Palm Bay's is fully invisible. Same state, same trade, same month. Reviews alone do not decide the outcome.

Zero agreement: the engines never named the same top plumber

In our June 2026 roofing measurement, the engines agreed on the same top pick in one city out of 16. Plumbing is worse. In the July 2026 run, the three engines' single most-named plumber was identical in zero of 16 cities, and the average city produced 2.6 different "#1 plumbers" out of a possible 3.

The engines also behave differently as instruments, measured across the same 640 answers each in July 2026:

Two patterns worth acting on. First, Gemini is the most decisive engine: its top pick appears in 78% of its own runs on average, so winning Gemini means being repeated almost every time, and losing it means near-total absence. Second, Grok leans on the national brand: Roto-Rooter is Grok's most-named plumber in 8 of the 16 cities we measured in July 2026, and Roto-Rooter shows up in at least one engine's top-10 list in 15 of 16 cities. When an engine is unsure who the best local plumber is, the franchise with a page for every city is the safe default. That is a beatable default, but only for plumbers who show up where the engines look.

The Brevard County story: four cities, four failures

The Space Coast is our home market, so here is the July 2026 measurement for all four Brevard cities in the panel. Each one fails a different way.

• Palm Bay: the review leader, 1-Tom-Plumber (1,871 reviews), was named in zero of the 120 recorded Palm Bay answers across all three engines. The engines do know the business; it surfaces in a handful of Melbourne answers in the same run. They just never connect it to Palm Bay, the city where it leads Google's reviews. A review leader that AI files under the wrong city is invisible to every Palm Bay homeowner who asks.
• Cocoa: Freedom Air and Plumbing holds 7,732 reviews, the second-largest count in our 16-city panel, and was named in 1 of the 120 recorded Cocoa answers in July 2026. That is a 1% share for the review leader in its own market.
• Melbourne: EK Coggin Plumbing (2,822 reviews) splits the engines three ways. In July 2026, Gemini named it in 36 of its 40 Melbourne answers and made it its most-named pick, Grok named it in 11 of 40, and Claude named it in zero of 40. Same business, same city, same week: dominant on one engine, absent on another.
• Titusville: EK Coggin also leads Titusville's reviews and reached 39% of runs there in July 2026, the strongest Brevard result. But both Grok and Gemini still most-named Roto-Rooter in Titusville, so the local leader with 2,822 reviews sits behind a national franchise on two of three engines.

For Brevard plumbers this is the whole pitch in one county: the businesses that won the Google reviews game have not won, and mostly have not entered, the AI recommendation game. The gap is measurable, and it is fixable, because the next two sections show exactly where the engines get their names.

Where AI looks: Yelp first, Angi second, then your own site

The assistants do not invent names; they cite sources. Across all 2,174 recorded answers in the July 2026 run, we logged 8,516 citations. Two engines hide their sources (Claude exposes no citations, and Gemini routes 4,727 citations through a Google proxy that masks the real domain), which leaves 3,789 citations from Grok and OpenAI that reveal a real source domain. Here is what they point to.

Behind the top five: thumbtack.com (70), facebook.com (56), Yahoo Local (40), HomeAdvisor (37), and BBB (26), plus a long tail of individual plumbers' websites. In the roofing edition Angi led the Brevard markets; for plumbing, measured in July 2026, Yelp leads statewide and it is not close.

Three takeaways from this table. First, Yelp and Angi are the two directories a Florida plumber cannot skip; they are cited more than everything else combined among directories. Second, rotorooter.com at #3 explains the franchise default: Roto-Rooter's own city pages feed the engines directly, which any local plumber with real city pages can also do. Third, Reddit at #4 means community threads ("who's a good plumber in Tampa?") are now a recommendation source, not just chatter.

What the winners have on their sites

Directories get you found; your own site is what the assistant reads to confirm you are real. We crawled the websites of the plumbers the engines recommend across all 16 markets in July 2026: 124 unique winner domains, of which 120 responded.

• 79 of 120 winner sites (66%) carry LocalBusiness structured data, measured July 2026.
• 95 of 120 (79%) show a phone number on the page.
• Only 18 of 120 (15%) expose their rating and review count in structured data.

Plumbing vs. roofing: the alignment is getting worse, not better

Because the plumbing run reuses the roofing battery (same 16 cities, same 8-question structure, same 5 repetitions), we can compare trades directly. To keep the panel identical, the roofing column below is recomputed on the same three engines and the same entity matching as the plumbing column.

(The published roofing report states 28% and 1 of 16 because it ran a four-engine panel; the numbers above restrict it to the same three engines for a fair comparison.)

On every alignment measure, plumbing is worse: the review leader surfaces less, and the engines disagree more. AI plumbing recommendations are even less tied to review reputation than AI roofing recommendations were a month earlier. Whatever is deciding these answers, it is not the Google review count, and a plumber waiting for their reviews to carry them into AI answers is waiting on a signal the engines are weighting less in this trade.

The playbook: how a Florida plumber earns AI recommendations

Methodology and honest limits

We would rather under-claim than over-claim, so here is exactly what this is and is not.

• What it is: measured primary data, collected July 29, 2026. A frozen battery of 8 plumbing questions, asked 5 times per engine per city across 16 Florida cities, recorded verbatim, with businesses, citations, and on-site signals extracted from those records and cross-referenced against a Google Places top-20 pull (309 listings) from the same month.
• The partial OpenAI panel: Grok (grok-4.3), Gemini (gemini-2.5-flash), and Claude (claude-haiku-4-5) completed all 640 planned answers each. OpenAI (gpt-4o-mini-search-preview) captured 254 of 640 before its account's prepaid credit ran out mid-run, covering roughly the first six cities. All headline cross-engine statistics use the three complete engines. The citation and winner-site layers pool all 2,174 recorded answers, including the partial OpenAI set, and are labeled that way where quoted.
• Entity resolution: business names in AI answers vary ("E.K. Coggin" vs "EK Coggin", curly vs straight apostrophes). Our first automated pass missed some of these variants and undercounted two review leaders; the numbers published here use corrected matching, verified by hand against the raw answer records. We say this because a skeptic checking our raw files against an older script output would find the discrepancy, and they should hear it from us first.
• Citation gaps we disclose: Claude exposes no citations, and Gemini routes its citations through a Google proxy that hides the real source domain (4,727 of the 8,516 logged citations). The where-AI-looks table is built from the 3,789 Grok and OpenAI citations that reveal real URLs.
• API surfaces, not consumer apps: we measured the assistant APIs with web search or grounding enabled. These are the best measurable proxies, but they are not identical to the consumer chat apps, and Google's AI Overviews have no API and are not measured here.
• No geo-spoofing: every prompt embeds the city in its text, and all runs execute from a single fixed location. The geographic signal comes from prompt wording, not IP localization.
• Coverage: 16 Florida cities, one trade, one month. A measured sample, not a statewide census. The method is what scales; this is the second industry edition, after roofing.
• Site crawl scope: homepage structured data and contact signals only, not full audits. 120 of 124 unique winner domains responded; the 4 that did not are excluded.

How to cite this report

The raw instrument data lives in the L3ad Solutions research archive (data/flsi/raw-instrument/v0.1/runs/2026-q3-plumbers-3e/ and data/flsi/rankings/plumbers-2026-q3-places.json in the site repository). It is licensed for free use with attribution. To cite this report:

L3ad Solutions Research. (2026, July 29). The State of Plumbing Search in Florida (2026). Retrieved from the L3ad Solutions research archive.

To save a printable copy, use your browser's "Print to PDF" function on this page.

Want to know where your business stands?

This report covers plumbing in 16 cities. The same measured check runs for any local trade in any Florida market, and we already run it for plumbing companies we work with.