How to get your business recommended by ChatGPT, Gemini, and Claude

Ben McDermott · September 3, 2026 · 4 min read

Type "best plumber in Minneapolis" into Google and you get ten links and a map. Ask ChatGPT the same thing and you get three names and a reason for each. If yours isn't one of them, you never knew you lost the job. Here's how the models decide, and what a small business can do about it.

How an AI assistant picks a business to name

When you ask ChatGPT, Gemini, Claude, or Perplexity a question about a local business, the model doesn't answer from memory alone. It runs a web search, reads a handful of pages, and writes an answer from what it found, usually with citations. Google's AI Overviews and AI Mode do the same thing on top of Google's own index.

That has two consequences. First, the sources are familiar: your website, your Google Business Profile, review sites, industry directories, local news, and the "best of" roundups that already rank on page one. Second, the model needs to pull facts from those pages quickly: what you do, for whom, where, at what price, and why someone would pick you. Pages that state those things plainly get used. Pages that make the model guess get skipped.

Start by measuring, not guessing

Before changing anything, write down the fifteen to twenty questions a customer would ask before hiring you. Not keywords: full sentences, the way people talk to an assistant. "Who does commercial snow removal in the west metro?" "How much should a small business pay for bookkeeping in Minneapolis?" Run each one through ChatGPT, Gemini, Claude, Perplexity, and Google AI Mode. Record who got named, what was said about you if anything, and which pages were cited.

Use a fresh session each time, because earlier conversation changes the answers. Expect variation between runs; that's normal, and it's why you rerun monthly instead of reacting to one result. This spreadsheet is your baseline, and the citations column tells you which sources to go after.

Fix your own site first

Say what you do in the first paragraph, in plain words, with the city. "1771 is a Minneapolis AI consultancy for owner-led businesses" is a sentence a model can lift. "Empowering transformation through innovation" is not.
Answer real questions on real pages. An FAQ that answers "what does it cost" with a number and "who is this for" with a description gives the model exactly what it needs to quote you. Put prices on the site if you possibly can.
Add structured data. Schema.org markup for your organization, address, hours, offers, and FAQs turns prose into facts a machine can read without guessing. Google's guidance is that there's no special markup for its AI features beyond what already helps search, so this is the same work done once.
Add an llms.txt file: a short plain-text summary of your business, pages, and pricing at yoursite.com/llms.txt. It's a 2024 proposal, not a standard the providers have formally committed to, so treat it as cheap insurance rather than the main event.
Make sure the crawlers can get in. OpenAI, Anthropic, Google, and Perplexity each run their own crawler with its own user agent. Some hosting firewalls and security plugins block them by default. Check your robots.txt and your host's bot settings, and unblock the ones you want.
Serve real HTML. If your site is a JavaScript app that renders blank until scripts run, some crawlers see nothing. Prerender or server-render the pages that matter. This site does exactly that.

Then go where the models are looking

Look at the citations in your baseline. In most local categories the same five to ten sources show up again and again: Google Business Profile, Yelp, the Better Business Bureau, an industry directory or two, a local publication's "best of" list, and a Reddit thread. Those are your targets.

Claim and complete every profile, with the same name, address, phone, hours, and description everywhere. Inconsistency reads as two different businesses.
Get reviews that say what you did, not just that you were great. "Fixed our furnace on a Sunday in January" is a sentence a model can use to recommend you for emergency furnace repair.
Get on the lists. If a local publication runs "best accountants in the Twin Cities," being on it matters more than it ever did, because that one page feeds a hundred AI answers.
Publish something worth citing. A page with real local numbers, a clear how-to, or an honest price comparison gets pulled into answers because it's the most useful thing the model found. Our article on what an AI consultant costs in Minneapolis exists for exactly this reason.

Rerun it monthly and write down what changed

Same questions, same tools, fresh sessions, once a month. Note where you appeared, where you dropped out, and which new sources showed up in the citations. Pick one thing to fix before the next run. This takes about an hour once the sheet is built, and it's the step most businesses skip, which is why the ones that don't skip it pull ahead.

What to be skeptical of

Anyone who guarantees placement in ChatGPT or Gemini. The answers change from run to run and the models change monthly. The honest promise is the work and the measurement.
Tools that report a single "AI visibility score." Useful for a trend line, meaningless as a number. Look at the actual answers to the actual questions.
Spending on "AEO" or "GEO" before the basics are done. If your site doesn't state your price and your city in plain text, no amount of optimization fixes that.

Sources

More insights

Book a free 30-minute call