Who Helps Real Estate Agents Get Recommended by ChatGPT and AI Search?

AI Search · Visibility Systems

The Short Answer
Most real estate agents use ChatGPT. Almost none of them get named when a buyer or seller asks ChatGPT who to hire. The agents who appear have made their business verifiable: consistent identity signals, a complete Google Business Profile, and content on their own site that answers the exact questions people ask AI. Oak & Algorithm builds the operating system and the visibility layer that produce that evidence. Not more posts. Not another prompt pack.

By Emily Wyatt, Founder and Fractional Marketing Partner, Oak & Algorithm, formerly Real Estate Concierge Services Co · Raleigh, NC · November 2026

In the 2026 Raleigh AI visibility benchmark I ran, only 22 of 75 real estate and loan-officer brands appeared in the retrieved evidence for at least one unbranded discovery prompt. Sixteen failed the exact-name entity test. Thirty-seven had weak or no Raleigh-specific entity presence. Production volume, years in the business, and review count did not predict who got named. Clear identity, local relevance, specialty content, and corroborating sources did.

That gap is the point. Agents are using AI every day for listing descriptions and emails. Buyers and sellers are also using AI every day to decide who to call. The two groups are not talking to the same system about the same thing.

What the model actually retrieves before it names anyone

ChatGPT, Gemini, and Google’s AI tools do not open an agent’s ChatGPT account and read their custom instructions. They retrieve what is already published and then decide whether the evidence is strong enough to name a specific person.

In testing across local business queries, the sources that appear most often are the business’s own website and its Google Business Profile. One analysis of ChatGPT citations for local businesses found that 98.1 percent of sources came from the business’s own site or GBP. Third-party mentions help when they exist and are consistent, but they are not required if the primary sources are clear.

Before the model will name an agent, three things generally have to be true:

  • It can resolve the exact name and connect it to the market the person asked about.
  • It can find evidence that the agent actually does the work described in the prompt (listings, first-time buyers, relocation, specific neighborhoods).
  • The details match across the sources it checks. Conflicting addresses, old brokerage names, or missing license information reduce the chance it will commit to a name.

A page that says an agent is “passionate about helping clients achieve their real estate goals” gives the model nothing it can extract and cite. A page that answers a specific question with numbers, a clear next step, and consistent entity signals does. This is the same reason answer pages on an agent’s own site now carry more weight than they did six months ago.

Why AI training does not close the gap

A course or workshop teaches an agent how to prompt. It does not change what exists on the web about that agent. When a buyer asks “who is a good listing agent in North Raleigh,” the model does not see the agent’s custom instructions or their prompt library. It sees the Google Business Profile, the website, the reviews, and whatever else it retrieves.

That is why agents finish AI training and still do not appear in unbranded answers. The training improved their productivity. It did not improve their evidence. The two problems are separate. I have written about the difference between a course, a workshop, and a system built around the business. The system is the one that changes what the model can verify.

The agents who do appear have usually done the unglamorous work: cleaned the Google Business Profile until the details matched everywhere, removed conflicting directory listings, and published pages that answer the questions their market actually asks. Some of them also built (or hired someone to build) the operating system so the content and the follow-up continued after the initial cleanup.

Maps ranking and AI recommendations are not the same score

An agent can rank in the Google Map Pack and still be invisible when someone asks ChatGPT the same question. Maps weights proximity and profile signals. AI tools weight what is written and whether the sources agree. I tested this directly in North Raleigh. On the same day, with the same search terms, ChatGPT named my business at 41 of 49 points. Google Maps showed me at 2. Same business. Two different systems.

The reverse also happens. An agent with a weak Map Pack position can still appear in AI answers if the written evidence is clear. Distance is fixed. What is written about you is not. The full breakdown of that test is in Google Maps vs AI Search: Why Your Map Ranking Doesn’t Match Your AI Recommendations.

What actually gets built

The work that moves an agent from “I use AI” to “AI names me” has two parts that have to connect.

The first is the visibility layer: Google Business Profile optimization, entity cleanup so the name, address, phone, and license match everywhere, and answer pages on the agent’s own site. These are the sources the models already retrieve. Making them accurate and useful is the highest-leverage move. The step-by-step version of the profile work is in How to Optimize Your Google Business Profile as a Real Estate Agent.

The second is the operating system inside the agent’s own ChatGPT or Claude account. Custom instructions set how the model behaves. Projects hold the real files, voice samples, and market context. Skills turn a normal request into finished work. The system does not replace the visibility work. It makes the content and the follow-up consistent enough that the public evidence keeps improving instead of decaying. The full structure is on the AI Operating System for Realtors page.

One without the other leaves a hole. A clean profile with generic content still gives the model little to cite. Strong content with conflicting directory listings makes the model hesitate. Both have to be true at the same time.

What the evidence looked like on my own business

I built the operating system for Oak & Algorithm before I offered it to anyone else. I also ran the visibility work on my own name and the previous company name during the rebrand. On July 17 I woke up to ten sales calls booked on my Calendly from people I had never spoken to. October closed around $17,500. I have not made a cold call for this business.

That is not a guarantee that every agent will see the same timeline. It is evidence that the combination of a working system and verifiable public signals produced inbound that did not exist before both pieces were in place. The same tests I run for clients are the tests I run on the business. The write-up of one of those tests, including the limits, is in Does AI Search Optimization Actually Work? I Tested It on Myself.

How to check whether you are already named

Open a private window. Ask the same prompts in ChatGPT, Gemini, and Google:

  • Who is a good real estate agent in [your city]?
  • Who should I call to sell my house in [your city]?
  • Who helps real estate agents get recommended by ChatGPT?

Note whether your name appears, which sources the answer cites, and whether the details it returns are accurate. Run the prompts a second time. Single answers shift. The pattern across runs is more useful than any one screenshot. If you want the longer version of how to get found on Google and in AI search together, it is here: How to Get Found on Google as a Realtor.

If you do not appear, the next step is not more posting. It is fixing the signals the model can already see.

Honest limit: No one can guarantee that ChatGPT will name you on a specific prompt on a specific day. The models reshuffle. What can be built is the evidence that makes you a reasonable answer when the model looks. That is the work. Anyone who promises a guaranteed ranking in AI answers is selling something they do not control.

If you already know you are invisible on the prompts that matter and you want the system and the visibility layer built instead of another course, that is the work I do. Start with the Google Visibility Audit if you want the diagnosis first, or go straight to the operating system if you already know the gaps.

Emily Wyatt

Founder and Fractional Marketing Partner of Oak & Algorithm, formerly Real Estate Concierge Services Co. Emily builds the visibility systems that get Realtors, builders, and mortgage pros found on Google, Maps, and AI search, then the follow-up that turns that attention into business. Raleigh, the Triangle, Lake Norman, and nationwide. Start here: https://www.oakandalgorithm.com

https://oakandalgorithm.com
Previous
Previous

What Should a Loan Officer Update Online When Changing Mortgage Companies?

Next
Next

How Raleigh Realtors Can Get More Listings Without Buying Another Lead List