Why AI search requires a framework, not tactics
AI interfaces, citation behavior and retrieval methods change quickly, so tactics tied to one platform expire. A framework lasts because it governs principles that stay necessary when platforms change: be understandable as an entity, make useful information extractable, build credible corroboration, and measure what actually happens.
The four components of the framework
The components run in a fixed order, because each depends on the one before it. An assistant can’t extract facts about a builder it can’t identify, and measurement only helps once there is something to measure.
Component one: entity definition
Establish consistent relationships among the organization, its markets, communities, sales centers, homes, leadership, structured data and authoritative third-party references, so every source describes the same builder. This is the work of entity and authority engineering.Component two: content extractability
Use descriptive headings, direct answers, self-contained explanations, tables, steps and definitions so the important passages still make sense when an AI system quotes them outside the full page.Component three: corroboration and citation sources
Strengthen credible evidence from local publications, earned media, official and municipal sources, industry organizations, reviews and business profiles. This is not mass directory or link acquisition.Component four: measurement and iteration
Track prompt outcomes, brand mentions, citations, share of answer, competitor inclusion and AI referral traffic over time, then use the findings to decide whether entity, content, corroboration or structure is the current constraint.
Structured data is the translation layer
Structured data tells machines what each page is about and how its pieces relate. When every visible element has an intentional structured counterpart, the site stops being a collection of pages and starts working as a knowledge graph about your company. It matters most where visitors and machines see different things: a price range or availability list that loads only after a click may be invisible to a crawler, and schema states it explicitly.
A schema mapping for builder websites
This mapping, adapted from Cristobal Varela’s book, What Executives Get Wrong About SEO, shows how common elements on a builder site translate into structured data.
| On the page | What the markup tells machines | Why leadership should care |
|---|---|---|
| FAQs | Questions and answers marked up as pairs, tied to the page’s main topic (FAQPage) | Gives answer engines clean question-and-answer pairs to quote in your own wording |
| Community location and sales center contact details | Location details marked up for maps and assistants (Place, PostalAddress) | Helps assistants and maps return the right address, phone number and hours for each community |
| Inquiry and tour calls to action | The next step a buyer can take, attached to the community or service (potentialAction) | Makes the next step, such as requesting information or scheduling a tour, discoverable to machines |
| Authors and experts | The real person behind the content, with credentials and profiles (Person) | Shows a real, qualified person stands behind the content, which AI systems weigh when choosing sources |
| Reviews and reputation | The company and its reviews, tied to real profiles (Organization, Review) | Connects the reputation buyers can verify elsewhere to the entity on your site |
| The page’s main content | What the page is primarily about: an article, a service, a home or a community (mainEntity) | Tells search engines and assistants what the page is primarily about, so it is matched and summarized accurately |
The right types depend on what each page actually shows. Markup should describe only content visitors can see, and never claim ratings, prices or availability the page doesn’t support.
Why we write schema instead of using plugin output
Schema should be custom-written to your business model: how your divisions, communities, floor plans and homes relate. Plugin output tends to duplicate data, misstate relationships or leave out the context machines rely on, and the wrong types can confuse search engines rather than clarify. Leadership should treat schema as governance, a commitment to accuracy, not decoration.
llms.txt and AI.txt are proposals, not standards
AI.txt, pitched as a robots.txt for AI, and llms.txt, a menu pointing language models to approved summaries, are both proposals, not adopted standards, and AI systems are not required to honor them. They can be explored experimentally, but robots.txt and server-side access rules remain the dependable controls for allowing or blocking crawlers, including AI-related ones. For a builder that wants to be named in AI answers, blocking those crawlers usually works against that goal.
How content is structured for extraction
We put the main answer at the start of each section, then the context and detail buyers need to evaluate it. Headings explain their topic without relying on the surrounding page, so people and AI systems can understand each section on its own. Where they help, we use comparisons, numbered processes and definitions, such as plan comparisons across a community or the steps from inquiry to contract.
How the framework is governed month to month
Prompt testing cycles
We test a recurring set of buyer questions by market to monitor brand inclusion, competing builders, citations and coverage gaps. Our guide to AI search performance measurement explains what these tests can and can’t prove.
Mention, citation and share-of-answer tracking
We measure where the builder is consistently named and described accurately and where it is absent or misrepresented, then work on defending AI mentions and citations where answers drift.
Content iteration based on findings
We use the findings to prioritize clearer entity signals, stronger corroboration or better content structure, depending on what is limiting visibility.
How this framework applies to home builders
We build market-level prompt sets around the questions buyers actually ask: the best home builders in a metro, builders with quick move-in homes, builders offering a home type, and new home communities near a location. Then we connect the builder with its markets and communities, make the relevant facts extractable, strengthen corroboration and measure inclusion over time. The framework is the method behind our AI search visibility service.
