# What 200+ Local Site Builds Showed Us About AI Search | AiSeoCourse.net > The lessons in this course come from LocusPilot URL: https://aiseocourse.net/blog/what-200-local-site-builds-showed-us-about-ai-search/ Last-Modified: 2026-09-20 Author: Adam Yong would turn main into a scroll container and break every \`position: sticky\` inside it. --> Skip to main content [#main-content →](#main-content) Case Studies # What 200+ Local Site Builds Showed Us About AI Search The lessons in this course come from LocusPilot's builds since December 2025 and their Search Console data. Here are the patterns that shaped the curriculum. person Adam Yong calendar\_today Published September 4, 2026 update Updated September 20, 2026 schedule 7 min read ![Two hundred local business site thumbnails arranged in a grid with Search Console trend lines overlaid](/images/featured/flat-editorial-illustration-of-two-hundred-local-b.webp) This course did not start as a course. It started as the same three answers, given over and over, during site builds for local businesses. Since December 2025, LocusPilot has built more than 200 of them: trades, clinics, salons, restaurants, professional practices. Along the way we watched their Search Console data, their server logs and how assistants answered questions about them. This post describes the patterns that came out of that, and how those patterns decided which lessons became the ones with the most supporting material. A note on evidence, because it matters. The 200+ figure is a count of sites built by one publisher since December 2025. It is not an industry study, it is not a controlled experiment, and nothing here should be read as a universal law. Where a pattern showed up repeatedly across different trades and regions, we treated it as worth teaching. Where it did not, we left it out. That standard is set out in the editorial policy [/editorial-policy/ →](/editorial-policy/) . ## Pattern one: the problem was almost never the website Owners usually arrived convinced they needed more content. In most cases the site was not the binding constraint. The binding constraint was that the facts about the business disagreed with each other across the web. ![Panel showing the recurring causes of invisibility across builds](/images/content/pattern-panel-showing-recurring-causes-of-invisibi.webp) An old phone number on two directories. A suite number written three ways. Hours that changed a year ago everywhere except the profile. When we asked an assistant about those businesses, the answers hedged: vague on hours, cautious about the phone number, sometimes naming a competitor with a duller website and a unanimous footprint. That is why Lesson 2 [/lessons/google-business-profile-for-ai-search/ →](/lessons/google-business-profile-for-ai-search/) became a priority lesson with five supporting guides. Cleaning up a footprint is unglamorous and it moves the needle more reliably than publishing. ## Pattern two: pages got skipped, not outranked The second recurring finding was about what happens when a site does get read. Assistants lift passages. A page full of confident adjectives and no specifics gives them nothing to lift. The pages that got quoted had a shape in common. A question as a heading, an answer in the first sentence, then a specific fact: a named service area, a timeframe, a price range, a piece of equipment. The pages that got ignored opened with a paragraph about a commitment to quality. Rewrites were usually cheaper than expected. Reordering existing sentences and replacing adjectives with numbers changed retrieval outcomes on several builds within weeks. That is the substance of Lesson 3 [/lessons/local-site-content-ai-can-cite/ →](/lessons/local-site-content-ai-can-cite/) . ## Pattern three: schema helped at the margin, not at the core We added LocalBusiness JSON-LD on effectively every build, so we cannot claim a clean comparison. What we can say is narrower and more useful: on sites where the business facts were ambiguous, such as a service-area business with no public address or a business sharing a name with another in the same region, structured data appeared to help engines resolve the entity. On sites where the facts were already clean and consistent, adding markup rarely produced a visible change on its own. So schema earns its place in Lesson 4 [/lessons/schema-and-structured-data-for-local-ai-seo/ →](/lessons/schema-and-structured-data-for-local-ai-seo/) with five supporting guides, and the lesson is explicit that it is not a substitute for having something worth quoting. ## Pattern four: measurement was the weakest link This was the uncomfortable one. Owners asked whether the work was landing, and the honest answer was that the tooling is immature. Search Console shows impressions where AI Overviews appear, but not cleanly separated from everything else. Some assistants send referral traffic that can be segmented, and some send none at all. Citations appear and disappear between runs of the same prompt. None of that is a reason to stop measuring. It is a reason to measure in a way that survives noise: a fixed set of customer-style prompts, run logged out, on a monthly cadence, recorded with dates. ![Panel mapping observed failure patterns to lesson priorities](/images/content/priority-mapping-panel-converting-observed-failure.webp) Lesson 7 [/lessons/measure-your-ai-search-visibility/ →](/lessons/measure-your-ai-search-visibility/) is built around that method, and it spends as much space on what cannot be measured as on what can. ## How the patterns became the curriculum Four lessons carry the most supporting guides because four problems came up the most: | Observed problem | Lesson | Supporting guides | | --- | --- | --- | | Owners did not know how AI answers were assembled | Lesson 1 | 6 | | Business facts disagreed across listings | Lesson 2 | 5 | | Structured data missing or wrong on ambiguous entities | Lesson 4 | 5 | | No repeatable way to tell whether anything worked | Lesson 7 | 3 | Everything else in the course exists because it was needed to make those four work: content shape in Lesson 3, the technical floor in Lesson 5, off-site signals in Lesson 6, and an honest exit in Lesson 8. ## What we still do not know Plenty. We do not know how long a corrected third-party record takes to propagate through every engine, only that it is slower than fixing your own site. We do not know how much weight any engine puts on review sentiment versus review count, only that specific reviews appear in answers more often than generic ones. Where the course states a limit like this, it states it as a limit. If you want the method rather than the findings, start with Lesson 1 [/lessons/how-ai-search-picks-local-businesses/ →](/lessons/how-ai-search-picks-local-businesses/) . If you want to check your own standing first, run the baseline test in Lesson 7 [/lessons/measure-your-ai-search-visibility/ →](/lessons/measure-your-ai-search-visibility/) and write down what you see. **See how the measurement actually works** — Read Lesson 7: Measure Your AI Search Visibility [/lessons/measure-your-ai-search-visibility/ →](/lessons/measure-your-ai-search-visibility/) ![Portrait of Adam Yong](/images/squares/editorial-headshot-portrait-of-course-author-adam-.webp) Adam Yong Author & Founder, LocusPilot Adam Yong is the founder of LocusPilot, an AI website builder for local businesses, and the author of this course. He has spent 20+ years in software, founded Agility Writer, and leads GEO strategy at ADE Marketing. verified Founder, LocusPilot and Agility Writer More about the author [/author/adam-yong/ →](/author/adam-yong/) ## More updates September 18, 2026 ### The Free AI SEO Course for Local Businesses Is Now Live [The Free AI SEO Course for Local Businesses Is Now Live →](/blog/ai-seo-course-for-local-businesses-now-live/) Open access · Immediate start ## Start the course today All 8 lessons, the 90-day checklist and the glossary are open. No signup, no paywall, no email capture. Start Lesson 1 (Free) arrow\_forward [/lessons/how-ai-search-picks-local-businesses/ →](/lessons/how-ai-search-picks-local-businesses/) View the 90-Day Checklist [/checklist/ →](/checklist/) No signup. No paywall. 8 lessons, ~80 minutes.