How AI decides which local business to recommend.

What we tested

We wanted to know how AI assistants actually choose which local business to recommend, so we ran the test directly. We asked a leading AI model, grounded in live Google data, to recommend the best local marketing and SEO help for a small business in Las Vegas, and we captured exactly what it did behind the scenes: the searches it ran and the sources it cited.

We paired that with citation data from a tool that tracks how often specific domains get mentioned inside AI answers. The goal was simple. Stop guessing how AI picks, and watch it pick.

Finding 1: AI does not answer your question, it explodes it

Ask an AI assistant one question and it quietly runs several. For our single Las Vegas query, the model fanned out into multiple background searches at once, including broader phrasings and per-industry versions like local SEO for plumbers and for HVAC.

This matters because your page does not need to match one keyword. It needs to be the best answer across a spread of related questions the AI generates on its own. Content built to answer a real question, not just target a keyword, is what gets pulled into that process. That is the core of answer engine optimization.

Finding 2: AI cites platforms, and YouTube dominates

We measured which domains actually get cited in AI answers across five local Las Vegas categories: roofers, HVAC, plumbers, solar, and local SEO. The same names came up every time, and they were not local business websites. They were platforms and aggregators: YouTube, Facebook, Reddit, Yelp, BBB, and Wikipedia.

YouTube was the single most-cited source in every vertical we tested. That is a direct signal: a short video answering a customer question can earn AI citations that your website alone cannot. The takeaway for a local operator is clear. Ranking on Google is necessary but not sufficient. To be recommended by AI, you have to exist on the platforms AI trusts, which is why winning Google and AI takes one connected strategy.

The data: who AI actually cites, by vertical

Here are the platforms Google AI cited most for each Las Vegas category we tested. Note what is missing from every row: individual local business websites. The platforms get cited, the operators do not.

• [x]Roofers: YouTube, BBB, Facebook, Wikipedia
• [x]HVAC: YouTube, Facebook, Reddit, Yelp, Instagram
• [x]Plumbers: YouTube, Facebook, Reddit, Wikipedia
• [x]Solar: Reddit, YouTube, Wikipedia, Business Insider
• [x]Local SEO: YouTube, Facebook, Eventbrite, Yelp

Finding 3: invisibility is the default

When we checked how often typical local businesses get cited in AI answers, the common result was zero. Most local operators have no tracked AI mentions at all. Their competitors get named and they do not, and they usually have no idea it is happening.

This is the opportunity hiding in the problem. Because so few local businesses have done the work, the ones that do stand out fast. The bar is low, and early movers get named while everyone else stays invisible.

What this means for your business

• [x]Use video — YouTube was the most-cited source in every vertical. Short videos answering common customer questions earn citations your website cannot.
• [x]Get into the platforms AI reads — an optimized Google Business Profile plus the review and social profiles (Yelp, BBB, Facebook) that feed AI answers.
• [x]Earn recent reviews — steady, specific reviews are among the strongest signals for both maps and AI.
• [x]Publish answer-first content — pages that directly answer buyer questions, so the AI fan-out finds and cites you.
• [x]Track it — measure your citations across ChatGPT, Claude, Gemini, Grok, and Perplexity so you know if it is working.

Methodology, data, and how to cite this

We believe in showing our work. We ran a Google-grounded AI model (Gemini 2.5) on a category query and captured its background fan-out queries and cited sources directly. Then we measured which domains actually get cited in Google AI Overviews across five Las Vegas verticals, roofers, HVAC, plumbers, solar, and local SEO, using the DataForSEO LLM Mentions API, ranked by mention count. It is a focused, transparent benchmark across six queries in one metro, directional rather than exhaustive, and we are expanding it to more verticals, metros, and platforms.

The aggregate, anonymized data is free to download and reuse: Las Vegas AI Visibility Test dataset (JSON, CC BY 4.0). To cite it: WebsitesOnDemand (2026). Las Vegas Local AI Visibility Test. Retrieved from https://websitesondemand.io/how-ai-recommends-local-businesses/

See where your business stands

You cannot fix what you cannot see. Our free audit scores your site across six areas and tests whether AI assistants recommend you today, then hands you a prioritized fix list in about five minutes. No card, no call required to read it, and the report is yours to keep.

If your business is one of the many with zero AI citations, the report shows you exactly why and what to change. Run the free audit and find out where you stand before your competitors do.