Getting recommended by ChatGPT, Claude and AI search

Something has changed over the last couple of years, quietly: plenty of people looking for an agent now start by asking an assistant. “Who’s a good realtor in Sarasota for waterfront condos.” “Find me an agent in Nashville who works with first-time buyers.”

Nobody has good numbers on how big this is yet, and we’d be suspicious of anyone who claims they do. But it isn’t zero, and the mechanics of how you get named are different enough from classic SEO to be worth understanding.

What decides whether an AI assistant names you?

Mostly, what other sources say about you. That’s the uncomfortable part, so it goes first.

When an assistant answers “good agents in Kansas City,” it isn’t reading everyone’s homepage and judging. It’s drawing on patterns across a lot of sources: directories, review platforms, roundups from local publications, forum threads, local news, and yes, the sites themselves. The sites are one input among many, and not the most trusted one.

A business that appears across a directory, a couple of local roundups, a forum thread and a news mention reads as real. A business that appears only on its own domain, saying excellent things about itself, reads as unverifiable. Assistants are, sensibly, cautious about the second kind.

This is annoying because it’s the part you control least. It’s also the part that matters most, so it’s better said plainly than buried under a checklist of meta tag tweaks.

Where should an agent be listed?

Wherever buyers and sellers check an agent, and wherever your market gets written about. Most of these are free. They usually pass little or no link value, which is fine, because what you’re building isn’t links. It’s corroboration.

Be consistent everywhere. Same business name, same spelling, same phone, same website. Matching one business across many sources is a real problem for these systems, and inconsistency makes you look like three half-documented businesses instead of one solid one.

Get into the roundups. The listicles that rank for “best agents in [city]” are disproportionately what gets quoted, because they’re structured and comparative. Many accept submissions, and few agents ask.

How should you get reviews?

Ask every client, make it easy, and never pay for them. Reviews with specifics are the ones an assistant can use.

Google’s review guidance suggests asking customers to “visit a Google link or scan a QR code.” It also says offering incentives “in exchange for customers to post reviews, change reviews, or remove negative reviews is considered fake & misleading content and is strictly prohibited.”

Federal rules go further. The FTC’s final rule on fake reviews, announced in August 2024, bans reviews “by someone who does not exist, such as AI-generated fake reviews,” and “compensation or other incentives conditioned on the writing of consumer reviews expressing a particular sentiment.” It also lets the FTC seek civil penalties against knowing violators.

What helps is substance. “Helped us buy in [neighborhood] after two other offers fell through” is a sentence an assistant can use. “Great experience!” is not. You can’t write the review for them, but you can ask the question that prompts a specific one: what was the hardest part, and how did it get solved?

What still helps on your own site?

Clear answers, structured data and specifics. It’s the half you can do this week.

Answer questions directly. Pages that state a fact plainly (what something costs, what’s included, who it’s for) get quoted. Pages that build atmosphere for six paragraphs before getting to the point don’t, because there’s nothing extractable in them. A short summary at the top of a page is worth more than it looks.

Structured data. Markup for your business, your location, your services and your FAQs tells a machine what a page is about instead of leaving it to infer. schema.org has a RealEstateAgent type for exactly this. It helps every kind of search, and we’ve covered the rest of the local search basics for agent websites. One caution: Google’s guidance on AI features says “there’s also no special schema.org structured data that you need to add” for AI Overviews. Markup helps understanding; it isn’t a secret key.

Specifics beat adjectives. “Serving the greater Sarasota area since 2011, with 40+ closings in Lakewood Ranch” is usable. “Your trusted local expert” is filler that appears on ten thousand sites and distinguishes you from none of them.

Put the basic facts in one place. Your about page should say, in plain sentences, your full name, your brokerage, where you’re licensed, the areas you work, the services you offer and the property types you handle, and how to reach you. It’s the page an assistant is most likely to check when it wants to confirm who you are, so make it easy to confirm.

An llms.txt file. The llms.txt proposal, first published by Jeremy Howard in September 2024, is a plain Markdown file at /llms.txt: a heading with your name, a short summary, and lists of links to your important pages. It helps when an assistant fetches your domain and wants a map. That’s a narrow but real case, it costs an hour, and it’s worth doing. No major AI search product has said it uses llms.txt to rank anyone, so it isn’t a growth strategy.

Can AI search tools actually reach your site?

Check, because some sites block them without meaning to. The rules are published.

Google. Its AI features guidance says “there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” A page has to be indexed and eligible to show with a snippet. The usual controls (nosnippet, max-snippet, noindex) limit what appears, so make sure they aren’t on pages you want quoted. Traffic from AI features is counted in Search Console’s Performance report under “Web.”

ChatGPT search. OpenAI’s crawler documentation says its OAI-SearchBot crawler is what indexes sites for ChatGPT’s search feature, and “sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers.” It’s a separate setting from GPTBot, which collects training data, so you can allow one and block the other.

Open your site’s robots.txt (yourdomain.com/robots.txt) and read it. If a security plugin or a template blocks every crawler it doesn’t recognize, you may be invisible to AI search without knowing it.

What mistake did we make on our own site?

We wrote our llms.txt like a sales document. It ran nearly a hundred lines, and about half of it compared us with competitors, each entry carefully explaining who they’d be a better fit for.

That’s a fine thing to have on a comparison page a person reads in full. In a summary file an assistant might quote three sentences from, we’d essentially written our competitors’ sales copy and handed it over. It also contained a price that was wrong for months, which is worse, because pricing is one of the things people most often ask an assistant about.

We rewrote it as a map: what we are, what things cost, and where the authoritative page for each subject lives. Shorter, and more useful.

The lesson generalizes. Whatever you write for machines should be the stuff you’d want quoted word for word, because that’s what happens to it.

How do you check where you stand?

Ask the assistants directly, then repeat it every quarter.

Three outcomes are possible. You’re named, in which case look at what gets cited. You’re not named but competitors are, in which case the gap is off-site. Or nobody local is named and it’s all national brands, which is the best case, because the position is open.

When an assistant gets a fact wrong about you (a price, a service, a location), the fix is correcting the sources it draws on, and you can’t fix what you haven’t noticed. Every one of our websites includes Google Business Profile setup, and Growth adds the local pages that give those sources something accurate to point to.