Most real estate agents now use AI tools regularly. Far fewer have checked whether those same tools ever mention them back. That gap, between using AI and being found by AI, is worth looking at with real numbers instead of guesswork.
What the confirmed data actually says
The National Association of Realtors' 2025 Technology Survey, based on a random sample of active Realtors, found that 68% have used AI tools in their business, with ChatGPT the most widely used tool by a wide margin. A separate 2026 survey of real estate professionals published by Realtors Property Resource found an even higher figure, 82%, among its respondents.
Those two numbers come from different samples and different years, so they shouldn't be read as one trend line. But they agree on the basic shape: a majority of agents have adopted AI tools into some part of their daily work, whether that's drafting a listing description, summarizing a contract, or answering a routine client question faster.
What that data doesn't say
Neither of those surveys measured whether AI tools recommend agents back. They measured whether agents use AI, not whether AI uses agents as a source. Those are different questions with different answers, and it's easy to conflate them.
A separate claim circulating in real estate marketing content says a large majority of agents are effectively invisible when an AI tool is asked to recommend one. That claim traces back to a benchmark report published by FlyDragon, a real estate AI marketing vendor, about its own study of AI citation patterns. It's worth naming the source plainly: this is a vendor's self-published research about the exact problem that vendor sells a fix for, not an independent academic or association study. That doesn't make the number wrong. It means the number should be read as one vendor's claim, not as an agreed-upon industry fact, and readers should go check the vendor's own methodology before repeating it as settled.
Why daily AI use doesn't translate into visibility
The two sides of the gap live in different systems, which is why one never feeds the other. When an agent types a prompt into ChatGPT, that interaction is private usage of a tool. When a buyer asks the same tool to recommend an agent, the tool answers from public sources: websites, business profiles, reviews, directories. Nothing about the first activity touches the second. An agent could be among the heaviest AI users in their office and have a public footprint that gives those same systems nothing to work with.
This distinction matters because fluency creates a false sense of coverage. Knowing how to prompt an AI tool feels like being "on top of AI." But the skill a buyer's question tests is different: it tests whether the public record of your business is clear, consistent, and specific enough to be quoted. Those are website and profile problems, and no amount of prompt skill fixes them.
Why the gap matters regardless of the exact number
Even without a single trusted figure for how many agents get cited by AI tools, the underlying mechanism is easy to reason through directly. An AI assistant answering "find me a real estate agent in [city]" is reading whatever it can find about agents in that market: their website, their Google Business Profile, reviews, and any other structured, citable source. An agent's site that has no clear service-area statement, no structured data, and no page that directly answers "who is this agent and where do they work" gives an AI system very little to cite, independent of any specific percentage.
That's a testable question on any individual agent's own site, not just an industry-wide statistic. Whether your name comes up says more about your one site than any survey does.
How to run the test on your own name
The test takes ten minutes and costs nothing. Open a fresh chat session with memory turned off, or use a browser where you're not signed in, so your own history can't leak into the answer. Then ask two kinds of questions.
First, the direct question: "Who is [your name], real estate agent in [city]?" This tests whether the system can resolve your identity at all. A wrong brokerage, an outdated service area, or a confusion with another person of the same name all point to weak or inconsistent public information.
Second, the category question: "Recommend a real estate agent for [property type] in [neighbourhood]." This is the question that produces new business, and it's much harder. You're no longer the subject of the question; you're competing to be part of the answer. Run both questions in more than one tool (Perplexity, Gemini, and Copilot draw on different sources than ChatGPT) and repeat them on different days, because answers vary between sessions. One absence proves little. A consistent absence across tools and sessions is a real signal.
Write the results down, even informally. A note of the date, the tool, the question, and whether you appeared turns next month's re-test into a comparison instead of a fresh guess. Agents who run this check on a regular schedule can see whether changes to their site moved anything, which is more than most of the industry can currently say about its own AI visibility.
What to check on an existing agent site
A few concrete things affect whether an AI system has something to cite:
- A clear "who and where" statement on the homepage or about page, stating the agent's name, brokerage, and service area in plain sentences, not just a logo and a headshot.
- Consistent information across the website, Google Business Profile, and brokerage directory listing. Conflicting details (a different office address on one page than another) give an AI system a reason to hesitate.
- Structured data that states the agent's identity and service area in a machine-readable format, covered in more depth in Saige's schema markup guide.
- Real, specific content about the agent's actual market, not generic copy that could apply to any agent in any city.
None of this guarantees a citation. It gives an AI system something concrete to work with instead of nothing. There is a useful way to think about the divide: the industry-wide percentage, whatever it turns out to be, is not actionable. Your own site's readability is. The first is a debate about methodology; the second is a checklist you can finish this week.
The takeaway
The confirmed data says most agents now use AI tools in their daily work. The claim that most agents are invisible to AI tools comes from a single vendor's self-published study and should be treated as that vendor's claim, not as verified fact, until it's checked against independent sources. What's worth taking from both, regardless of which numbers hold up over time, is that using AI and being found by AI are separate skills. An agent can be fluent in the first and have never tested the second. The test is free, it takes ten minutes, and it turns the whole debate into a fact about your own business instead of a statistic about everyone else's.
Related: Tracking whether AI assistants mention you · What is AEO for real estate agents · AI SEO vs. traditional SEO for real estate agents · How a Saige website rebuild works · Real estate website FAQ
