How AI Decides Which Roofer to Recommend

When a homeowner types "who should I call for a roof replacement in Austin" into ChatGPT or Perplexity, the AI does not flip a coin. It runs through a set of evaluation criteria and returns the sources that score best. Those criteria are documented well enough that you can deliberately optimize for them.

The contractors who appear in AI answers are not there because they paid for placement. They are there because their websites are structured in a way that AI can read accurately, evaluate as relevant, and cite with confidence. The ones who are not there are missing one or more of the same signals. Understanding the signals is step one.

The content match signal

The most direct driver of AI citation is page relevance. If a homeowner asks "how much does roof replacement cost in San Antonio," the AI looks for pages that specifically address that question, in that city. A page titled "Roof Replacement Cost in San Antonio" with a direct, city-specific answer is a citation target. A general services page that says "we offer roof replacement across the greater San Antonio area" is not.

This is the signal that most contractors are missing entirely. Their websites describe their services. They do not answer homeowner questions. Those are two different things, and AI engines treat them differently.

Each dedicated answer page you create is a separate citation target for a specific query. A contractor with 10 structured answer pages across their service area has 10 opportunities to appear in AI responses. A contractor with zero answer pages has none, regardless of how long their Google history is.

The technical structure signal

AI engines do not read your marketing copy and evaluate it the way a human recruiter would read a resume. They parse structured signals. The most important of these is schema markup: blocks of JSON data embedded in your page code that tell the AI directly what your business is, where it operates, and what it does.

A LocalBusiness schema with accurate service area data tells the AI your exact cities. A Service schema linked to your roofing work tells it what specific jobs you do. A FAQPage schema on your answer pages makes each question-answer pair directly extractable without the AI having to infer meaning from prose.

Without schema, the AI infers. Inference is error-prone. It gets service areas wrong, conflates similar business names, and often defaults to a directory that has the data encoded cleanly rather than a contractor site where the AI has to work for the information.

The authority signal

Beyond content and structure, AI engines factor in how credible a source appears based on how other trusted sites reference it. Aggregators like Angi have accumulated inbound references from local newspapers, Reddit threads, city guides, and consumer publications for two decades. That reference history signals credibility to AI engines.

Building authority for your own domain means earning mentions in those types of sources: local news coverage after a major storm, a city home-improvement guide that lists local contractors, a neighborhood forum thread where someone recommends your business. These accumulate slowly. They also compound. A contractor who has been mentioned in three local sources is more authoritative than one who has been mentioned in none, and the AI's citation rate reflects that difference.

The specificity advantage

For local contractor searches, AI engines favor specific over generic. A page that answers "how much does roof replacement cost in Dallas" with Dallas-specific figures and contractor context outperforms a national cost guide for that query, even if the national guide has much higher domain authority overall. The specificity of the match compensates for the authority gap.

This is the window that currently exists for independent roofing contractors. Angi is generic by design. It cannot publish a page that speaks specifically to post-hailstorm roofing in a single zip code with a specific contractor's name and experience. A local contractor's website can. For hyper-local queries, that specificity wins.

How the signals compound over time

These four signals do not operate independently. They build on each other in a sequence that determines how quickly your citation rate grows.

Answer pages and schema are launch conditions. You can have them built and live within 30 days. AI indexing typically happens within 30 to 60 days for active domains. Early citations begin appearing for your most specific queries first.

As citation history accumulates, the credibility signal strengthens. AI engines note which sources they have cited before for related queries and weight them accordingly. The contractors who build citation history earliest in their markets benefit from this compounding effect most.

Authority from external references builds on top of that foundation. Month 5 and beyond is where contractors who have both strong on-site structure and growing off-site mentions start appearing for broader queries that previously defaulted entirely to directories.

For a deeper look at the content side, read what homeowners ask AI about roofing and which answer pages to build. For the technical side, see how schema markup works for roofing websites. To understand why the gap exists in the first place, read why Angi shows up in AI answers instead of your site.