Why search is changing from "ranking" to "recommending"
Think of customer search in two different ways:
Type 1: Branded Search
The customer already knows your name (e.g. "Atta-Boy Plumbing Charleston"). They just need your phone number or website. This is easy, and 94% of companies pass this test.
Type 2: Recommendation Search
The customer has a problem, but doesn't know who to call (e.g. "Recommend a good roofer in Columbia, SC"). They ask AI or Google to recommend a company.
SC Marketing & AI tested 100 verified South Carolina home-service businesses (in HVAC, plumbing, roofing, and electrical) across 25 real-world recommendation questions. We compared live Google results with an OpenAI search-backed retrieval test.
The clearest finding was simple: being easy to find when someone already knows your brand name does NOT mean an AI search system will find or retrieve your business when a homeowner asks for a recommendation.
The Big Gap
In our recommendation search test, only 38% of the relevant trade businesses surfaced by official website in the AI search retrieval layer. Meanwhile, 94% of those same businesses ranked in Google's top 10 when searched by company name.
This shows a real gap: a business can look great on Google when people search its brand name, but be completely invisible when new customers ask AI tools who to hire.
About this research
SC Marketing & AI conducted this study to understand how South Carolina home-service businesses appear across branded search and recommendation-style discovery. The goal is not to crown winners or publish an AI ranking list. It is to give business owners a measurable way to think about discoverability as search behavior changes.
The research was collected on August 8, 2026. Because search and AI retrieval change over time, the findings should be read as a dated snapshot, not a permanent score.
Methodology
The research group contained 100 verified South Carolina home-service businesses drawn from seven markets: Greenville/Upstate, Columbia/Midlands, Charleston/Lowcountry, Myrtle Beach/Grand Strand, Florence/Pee Dee, Rock Hill/York County and Aiken. The group was assembled from verified outreach research and is not a probability sample of every home-service business in South Carolina.
For the branded search layer, each company was queried in live Google results using the format '[Business Name] South Carolina.' The study recorded whether the exact official domain appeared in the first 10 organic results, whether a matching local/entity feature was returned, and how many unique non-official domains appeared in the branded top 10. Google result data was collected through DataForSEO using desktop, English-language, United States settings.
For the recommendation-search layer, the study used 25 standardized prompts across HVAC, plumbing, roofing and electrical, covering 50 relevant businesses in the research group. The prompt format was 'Recommend a [service] in [city], SC.' A business counted as surfaced in the AI retrieval test only when its official domain appeared in the returned search-backed result set. The identical query strings were also run through conventional live Google results for comparison.
Important limitation: the AI retrieval test measured whether official domains entered a search-backed result set. It did not measure a personalized consumer ChatGPT answer, a Google AI Mode answer or a Perplexity answer, and retrieval is not the same event as a final generated recommendation. Results can change with time, location, personalization, model behavior and query wording. No causal claim is made that any specific source or mention produces AI visibility.
