Skip to main content
AI SearchAugust 24, 202610 min read

Gemini Local Search: What Contractors Need to Know

Gemini cites local business websites often, but its recommendations can shift between identical searches. Contractors need an owned-site strategy and a repeatable test, not a one-screenshot AI ranking report.

HL

Hunter Lapeyre

Founder, Obieo & Lapeyre Roofing

Gemini local search cited a business's own website for roughly 60% of the 14,472 citations in a July 2026 local-services study. That is a strong reason to keep service and location pages accurate. It is not proof of a Gemini ranking factor. Repeated tests also found that sources and recommended businesses changed often, so one result cannot show stable visibility.

For a roofer, HVAC company, plumber, electrician, pest-control company, or cleaner, the business decision is straightforward: strengthen the pages and local facts you control, maintain credible third-party proof, and measure visibility as a pattern. Then ask the only question that pays the crew: did the work produce qualified calls, estimates, and booked jobs?

The owner takeaway
Your website matters in Gemini's local answers, but a citation is not a lead and one answer is not a rank. Build the owned asset, repeat the test, and keep revenue proof separate from AI visibility.

What Did the Gemini Local Citation Study Find?

On August 6, 2026, local-search agency Steady Demand published its AI Citation Ledger. The team ran 1,487 Gemini queries across 50 U.S. metros and ten local-service categories. The set included plumbers, roofers, HVAC contractors, electricians, locksmiths, pest-control companies, and house cleaners, along with three other local categories.

The questions used three commercial templates, including “best” and “top rated” searches. The study logged citations from Gemini's Google Search grounding metadata and extracted businesses named in the answers. Its published methodology also documented the failed queries, classification rules, and comparison controls.

Local queries

1,487

Across 50 metros and 10 service categories

Citations logged

14,472

From Gemini grounding metadata

Own-site share

~60%

Directional; the study pages report 58.7% to 59.9%

Business websites were the largest source category

Roughly six in ten citations pointed to the local company's own website, more than directories, forums, review sites, and other source categories combined. Search Engine Land's independent coverage of the study highlighted the same practical result: the business website still supplies much of the supporting detail in Gemini's local answers.

That does not mean “publish more pages and Gemini will recommend you.” The original researchers explicitly note that Gemini may choose a business first, then cite its website to verify services, hours, or pricing. Citation share describes the displayed sources in this dataset. It does not reveal a causal ranking formula.

Home-trade source mixes were not identical

Plumbing, HVAC, electrical, and pest-control answers drew from a mix of owned sites, general directories, local-service directories, and community sources. That weakens any universal checklist. A citation source that matters for a roofer in one metro may contribute little to a cleaner or electrician in another.

What Does the Gemini Study Not Prove?

The study is useful because it publishes a query matrix, methodology, controls, and raw data. It is still vendor-led research from an agency that sells local-search services. A careful independent methodology review found three limits contractors should keep attached to the headline.

The evidence showsIt does not show
Which domains Gemini cited in this API-grounded query setA permanent citation share in the consumer Gemini app
Which businesses the extraction pattern recognized in the answersEvery recommendation Gemini made or why it made one
How results varied across selected prompts, metros, and roundsA local ranking factor, conversion lift, or booked-job result

The July 27-28 collection used gemini-flash-latest through the Gemini API with Google Search grounding. It did not scrape a logged-in consumer session, and the query carried the city and state as text instead of a device location. The business-name parser recognized roughly 70-75% of answers. Those constraints do not erase the pattern; they define where it applies.

The own-site number also moved as the researchers audited domain classifications: one page reports 58.7% after a correction, while the dashboard shows 59.9%. “Roughly 60%” is fair. Quoting a more precise figure would overstate agreement inside the source itself.

Confirmed vs. interpretation
Confirmed: owned business websites were the largest citation category in this Gemini test. Interpretation: clear service and location pages are worth improving because they give search systems accurate supporting facts. Unproven: that any single page edit will cause a Gemini recommendation or a booked job.

Why Is One Gemini Local Result Not a Ranking?

The main dataset compared three different phrasings of the same local need and found about 40% overlap among cited domains. The researchers then ran a smaller identical- prompt experiment: six fixed questions, 72 calls, and four rounds. Back-to-back citation overlap was 46.3%, and the model's own follow-up search strings overlapped only 5.6%.

A separate five-round control made the business impact clearer. Across pairwise round comparisons, Gemini returned the same top recommended business 7.9% of the time. A conventional Google local-pack control returned the same top listing 90.2% of the time. Those percentages measure repeat consistency, not the chance that any one contractor will rank.

Repeated local resultSame top businessWhat to report
Gemini recommendation7.9% of pairwise comparisonsA distribution across repeated checks
Google local-pack control90.2% of pairwise comparisonsA comparatively stable ordered result

The original grounding-drift experiment suggests the variation begins when Gemini generates fresh search queries, not only when web results change. Independent research from SurfacedBy found the same broad problem across several assistants: its repeated buyer-question study measured 48% citation turnover for Gemini and warned that changing sources do not always mean the top recommendation changed.

That distinction protects the budget. A vendor screenshot showing your company once is an observation. A month of repeated checks can show a visibility pattern. Neither becomes revenue until a homeowner visits, calls, qualifies, receives an estimate, and books.

What Should Contractors Change for Gemini Local Search?

Do not build a separate “Gemini website.” Google's official guidance for AI features in Search says foundational SEO still applies: make important pages crawlable, indexable, helpful, reliable, and eligible for a normal Search snippet. Gemini's consumer app and Google Search AI features are not identical products, but the safe operating work is the same work that strengthens the company's owned search asset.

Five changes to make first

1

Fix revenue pages before publishing more posts

Confirm each high-value service has a useful page that names the service, service area, common job conditions, process, proof, and next step. A roofing company's local search foundation should make repair, replacement, storm, and service-area facts easy to verify.
2

Make business facts agree everywhere

Match the website, Google Business Profile, major listings, licenses, hours, phone number, service areas, and categories. Use the Google Business Profile AI Mode audit to check the local information a homeowner may see before clicking.
3

Replace generic claims with verifiable proof

Show truthful project details, warranties, financing terms, service limitations, staff or crew information, and customer proof where appropriate. Do not create fake city pages, fake reviews, or unsupported “best contractor” claims to chase an AI answer.
4

Earn category-specific third-party evidence

Maintain the reputable directories, manufacturers, associations, community sources, and local publications that matter in the trade and metro. The study found no single source mix for every category, so audit observed sources before buying a broad listing package.
5

Keep technical SEO boring and correct

Return successful status codes, use descriptive titles and headings, publish consistent canonical URLs, keep internal links crawlable, and make structured data match the visible page. Our Google AI search guide for contractors explains which fundamentals deserve budget and which GEO shortcuts do not.

How Should a Contractor Test Gemini Local Visibility?

Treat Gemini visibility like a sample, not a position number. The goal is to learn whether the company repeatedly enters the consideration set for commercial questions in its real markets and which accurate pages or third-party sources support those appearances.

Four checks for a repeatable test

1

Choose questions tied to real jobs

Start with five to ten prompts for the services and locations closest to gross profit: emergency repair, replacement, inspection, financing, and comparison questions. Avoid broad educational prompts that cannot reasonably produce a local opportunity.
2

Hold the test conditions

Save the exact prompt, date, model or surface, location context, and account state. Test the same set on the same schedule. When conditions change, label the change instead of blending it into the old baseline.
3

Repeat each prompt

One run can be lucky or unlucky. Repeat the identical wording, then test a small set of natural variations. Record named businesses, order, citations, and which business facts appear. Report the share of checks where the company appears, not “rank No. 1 in Gemini.”
4

Compare engines separately

Gemini and ChatGPT cited the same domains only about 8% of the time in the original matched-query snapshot. Keep each engine in its own column. A blended AI score can hide the exact surface that gained or lost visibility.

If the team needs a broader prompt and entity baseline, use the AI-search brand visibility audit without turning its output into a fabricated market share. The useful output is a repeatable list of gaps and tests.

How Do You Connect Gemini Visibility to Booked Jobs?

Keep four evidence layers separate: appearance, visit, lead, and revenue. A citation proves only that a source was displayed in the checked answer. A referral proves a click reached the website. A qualified lead requires call, form, and CRM evidence. Revenue requires a booked or collected job under a consistent definition.

  1. VisibilityPrompt, location, engine, company mention, order, citations, and sample size.
  2. Observed visitReferral source, landing page, session, and conversion event when available.
  3. Qualified leadService fit, location fit, contact validity, appointment, and estimate status.
  4. Business resultBooked job, chosen revenue measure, gross profit, and evidence confidence.

The AI-search lead tracking guide shows how to keep Search Console visibility, AI referrals, calls, CRM records, and booked revenue in the right columns. If a homeowner says Gemini influenced the decision but no referral exists, record customer-reported influence. Do not rewrite it as observed referral attribution.

The operating rule
Continue the work when it fixes a known information gap, increases defensible visibility across repeated checks, or produces better qualified opportunities. Stop when the only proof is a screenshot, a citation count, or a proprietary score with no path to pipeline data.

Gemini Local Search FAQ

What does Gemini cite for local searches?

In a July 2026 study of 14,472 Gemini citations, roughly 60% pointed to a local business's own website. That was the largest source category. The result shows where Gemini found supporting information in this test; it does not prove that a website citation caused the business recommendation.

Does a Gemini citation mean the contractor was recommended?

No. A citation is a source attached to an answer. A recommendation is a business named in the answer. Either can change without the other, and neither proves a website visit, qualified lead, or booked job. Contractors should report each stage separately instead of calling every citation a win.

How often do Gemini local search results change?

The study found about 40% cited-domain overlap across three phrasings of the same local question. In a smaller identical-prompt test, back-to-back citation overlap was 46.3%. A separate five-round control found Gemini repeated the same top business in 7.9% of pairwise comparisons, versus 90.2% for Google's local pack.

Should contractors focus on their website or directories for Gemini?

Start with the website and Business Profile because they are owned or directly managed assets, then maintain accurate listings and real community proof. The study found different source mixes by trade and metro, so it does not support buying every directory listing or treating one platform as a universal Gemini tactic.

How should a contractor measure Gemini local visibility?

Use a fixed set of commercial prompts, repeat each prompt, record the date, location, named businesses, and citations, and compare the distribution over time. Keep that visibility report separate from referral visits, qualified calls, appointments, estimates, booked jobs, revenue, and gross profit in the company's CRM.

HL

Hunter Lapeyre

Hunter owns Obieo, a search and lead generation agency for home service companies, and Lapeyre Roofing. He reviews marketing changes by their operating result: qualified opportunities, booked work, and less dependence on rented demand.

Free Strategy Call

Turn AI Visibility Into a Measurable Local Pipeline

We'll review your local search footprint, revenue pages, tracking, and prompt baseline, then show you which gap is closest to qualified calls and booked work.

Book a Strategy Call

20 minutes. We'll be direct about fit.