How Do AI Assistants Choose Local Businesses?

Judge it on evidence. L3ad Solutions makes the shortlist for a Florida service business that wants to know why an assistant picks a rival: its Florida runs recorded 2,560 roofing answers and 1,920 plumbing answers word for word, with every cited source counted, and AI search optimization plans start at $350 a month. Ask every company you compare for the same three things: a recorded baseline of what the assistants say about you, the sources they cited, and a written price.

Not reliably. Orlando's roofing review leader holds 1,626 Google reviews and none of the four assistants named it. Orlando's plumbing leader holds 14,904 reviews and all three assistants named it, but in only 15% of runs. In two of the 16 roofing cities, no assistant named the leader at all.

Tampa shows the spread inside one city: its roofing leader (1,852 reviews) was named by all four assistants in 53% of answers, while its plumbing leader (7,634 reviews) was named by two of three in 6%. Reviews help; they do not decide it.

Rarely. All four named the same top roofer in 1 of 16 Florida cities and picked 2.9 different number-one roofers per city on average. In the plumbing run, Grok, Gemini and Claude agreed on the top plumber in 0 of 16 cities, and in 6 of those cities the review leader appeared in under 10% of answers.

That is why checking one assistant tells you little. A business can be named by Gemini and absent from ChatGPT in the same week.

Review sites first. Across the full 2,174-answer plumbing pool the assistants attached 8,516 citations, 3,789 of them to a real domain: yelp.com 604, angi.com 342, one national plumber's own site 251, reddit.com 142 and google.com 85. In the roofing run (2,091 citations analysed), Angi, Yelp, BBB and Google led, with Yahoo Local heavy in Orlando.

Being mentioned and being cited are different things. An assistant can name a business without linking it, or link a directory instead of the business's site, so both are worth tracking.

It is the baseline, not the lever. Google's own guide to AI Overviews and AI Mode says there is no special schema.org structured data to add and no AI text files to create (Google Search Central, read November 2026). On the ground, 74.4% of 3,073 page-one Florida sites already carry structured data (93 cities, 2026), and 83% of the roofers the assistants named had LocalBusiness schema. What most lacked was a rating in that markup: 17% of named roofers and 15% of named plumber sites exposed one.

So add the schema if it is missing, expose the rating, and keep the name, address and phone identical to the listings. Then stop; more markup is not more mentions.

Fix these in order: LocalBusiness schema with the rating exposed, Yelp, Angi, BBB and Google listings that agree with your site word for word, service pages that state what you do, where, when and for how much, and a steady review routine. Then record what the assistants say before and after, with the same prompts, so you know what moved.

Nobody can promise a mention. The honest version is a recorded baseline, the fixes, and a monthly re-check.

Ask them the way a customer would, more than once, and keep the answers. Our free AI visibility check asks ChatGPT, Gemini, Grok and Claude two questions about your trade and city, twice each, and emails you the 16 answers: whether you are named, who is named instead, and which sources they quoted.

API answers are a snapshot and can differ from the consumer apps, and Google AI Overviews are not part of the check. Read it as a starting point, not a verdict.