The AI Visibility Audit for Real Estate Agents: 7 Checks You Can Run Today

Seven checks, twenty minutes, no tools. Run them in order and you will know which signal is breaking.

By Andrew Hong. Last updated 11 September 2026. We re-run and refresh these checks quarterly.

A past client gives your name to a friend who is moving. Before that friend calls you, they type your name into ChatGPT.

Here is what usually comes back for an established agent. A two-line summary assembled from a directory. A brokerage you left in 2023. A specialty that belongs to another agent with a similar name. Sometimes nothing at all, and a polite suggestion to check Zillow.

That is not the catastrophe it sounds like. NAR's 2025 profile of home buyers and sellers puts 43% of buyers finding their agent through a referral and another 18% using an agent they had worked with before. Six in ten introductions still come from a person. What changed is the gap between the introduction and the call. The assistant is not replacing the referral. It is checking it.

So the question worth answering is not whether AI is sending you business. It is whether AI confirms the story the referral already told.

You can find out in twenty minutes. Seven checks, no tools, no login, nothing to buy. Run them in order and you will know which of your signals is breaking and what it takes to fix.

What AI is doing when someone asks about you

An assistant answering a question about a named professional is doing something closer to fact-checking than to searching.

It gathers what it can find: your website, your Google Business Profile, your Zillow and Realtor.com profiles, your LinkedIn, the brokerage roster page, whatever press or association listing exists.

Then it looks for agreement. Where the sources agree, it states the fact plainly. Where they disagree, it hedges, which reads to a buyer as uncertainty about you, or it moves to an agent it can describe with confidence.

That second behavior is the one that costs you. The assistant is not judging who is better. It is judging who is clearer. An agent with half your volume and one consistent story across five profiles gets named ahead of you, because naming you would mean choosing between three versions of your career.

The scale is modest and it is not staying that way. Bank of America's June 2026 Homebuyer Insights Report found 20% of prospective buyers and current homeowners used AI tools or chatbots in the past year while researching a home, rising to 32% of Gen Z. Cotality's 2026 data, reported by NAR, found 44% of buyers would pay more for a human to verify what AI told them. The assistant is the first read. You are still the decision, if the first read holds up.

Which is why the fix is rarely a single edit. Your story lives in six or seven places you do not think of as marketing, and a contradiction in any of them is enough to make an assistant cautious. We make the full case in our piece on why AI does not recommend you.

One thing our own search data shows: across three months, the questions agents ask AI about hiring help are full sentences, not keywords. The most common one is a request for an audit. So here is one.

The seven checks

Each check takes two to four minutes. Do them in order. The later ones make more sense once the earlier ones have failed.

1. The referral check

Open ChatGPT, Gemini and Google AI Mode. In each, ask the same question: "Tell me about [your full name], real estate agent in [your city]." Ask it cold, in a fresh chat, signed out if you can, so you see what a stranger sees.

A pass: the correct brokerage, the correct market, and a specialty you would claim out loud in a listing presentation. Nothing invented.

A fail: no useful answer, or a confident answer with wrong details. The wrong brokerage is the most common. A specialty belonging to someone else with your name is second.

What a fail means: the sources an assistant reads about you disagree, or they are too thin to support a description. A blank answer is a coverage problem. A confidently wrong answer is a consistency problem, and it is the more urgent of the two, because the buyer reading it has no reason to doubt it.

2. The short-list check

Same three engines, different question, and this time your name stays out of it. Ask: "Who are the best [luxury, waterfront, relocation, first-time buyer] agents in [your city]?" Run it once in each engine, then again a day later, because answers move.

A pass: you are named in at least one of the answers.

A fail: six answers, no mention, and a list of the agents you compete with.

What a fail means: you are not in the set an assistant can verify for that specialty in that market. Notice who is, then notice where those names appear that yours does not. Usually a local press mention, an association page, a podcast, or a best-of list you never knew existed. The check is worth more than its answer, because what it surfaces is your work order.

3. The story check

No AI in this one. Open four tabs: your website About page, your Zillow profile, your Realtor.com profile and your LinkedIn. Read them side by side.

A pass: the same specialty, the same market, the same brokerage and the same years in business in all four.

A fail: three different stories, which is the usual result. A luxury specialist on the website and a generalist on Zillow. Nine years in one bio, twelve in another. A brokerage you left, live on a profile you forgot you had.

What a fail means: this is the most common failure on the list and the cheapest to repair. Every one of those pages is a source an assistant reads. Pick the true version, write it once, and make the other three match it. Our guide to the bio AI quotes back has the template.

4. The name check

Search your business name in quotation marks in Google. Then search your phone number on its own. Read three pages of results and count the distinct versions of your name, your phone number and your office address that are live right now.

A pass: one version of each.

A fail: more than one of anything. Two phone numbers. A suite number written three ways. A team name you retired two years ago.

What a fail means: every variant is a separate entity as far as a machine is concerned, and your credibility is split across all of them. Reviews attached to the old listing do not count toward the new one. Unglamorous work with a real payoff, and name and citation cleanup covers how to run it.

5. The proof check

Take one page from your site, your About page or a listing page, and run the URL through Google's Rich Results Test. It is free and takes about thirty seconds.

A pass: structured data describing you as a person or an organization, plus your reviews and your listings, detected without errors.

A fail: nothing detected, or a generic page entry and nothing else. That is the result for most brokerage templates and most agent website platforms.

What a fail means: you are asking every machine that reads your site to infer who you are from paragraphs. Some will, many will not, and none as reliably as they read a labeled fact. Structured data is not a ranking trick. It is the difference between a fact an assistant can quote and a sentence it has to interpret.

6. The source check

Go back to the engine that gave you the best short-list answer in check two. Ask a follow-up: "What sources did you use for that?" Most engines will list them, or show citations inline.

A pass: at least one source you appear on.

A fail: portals, brokerage directories and best-of lists with no trace of you on any of them.

What a fail means: this is the check that tells you where the work is. Assistants build who-is-good-at-this answers out of third-party sources far more than out of the agent's own site. Your site is where an assistant confirms a claim. Someone else's page is where it finds the claim. If your name is on none of the sources feeding the answer in your market, site work alone will not move it.

7. The website check

Open your site, view the page source, and use find to search for your neighborhoods, your specialty and a listing address. Or turn JavaScript off in your browser settings and reload the page.

A pass: your bio, your market and your listings are in the HTML, readable as text.

A fail: an almost empty page. A header, a footer, and a container waiting for an IDX widget to load.

What a fail means: the page a person sees and the page a crawler reads are two different documents, and everything that makes you findable arrives after the read. This is common on agent website platforms, it does not show from the front end, and it is the most expensive failure on this list, because no amount of profile work compensates for a site with nothing in it. We take it apart in a site that looks right and reads empty.

The scorecard

Scoring it

Count the passes. Seven checks, seven possible.

Six or seven. Your signals are in order and this is maintenance. Re-run the checks quarterly, update the profiles whenever anything changes, and spend your attention on the source check, which is the one that keeps moving.

Three to five. One signal is broken and it is doing most of the damage. In the audits we run it is almost always the story check or the name check. Both are fixable in a week, and both pay off across every other check, because they give the remaining sources something to agree with.

Zero to two. There is no verified record of you for an assistant to draw on, and the cause is usually the website. Checks five and seven are the tell. When those two fail together, start there, because the profile work will not hold without it.

One note on reading your score. A low number is not a verdict on your business. Agents at every production level score badly here, including some of the best-known names in their markets, because these signals have nothing to do with how well you sell. They have to do with whether the record agrees with itself.

What to fix first

Order matters here more than effort does. Each fix gives the next one something to confirm against, so working out of order means doing it twice.

1. The story. Decide the true version: market, specialty, brokerage, years, credentials. Write it once. It is the source document for everything else, and until it exists the rest is guesswork.

2. The profiles. Push that version to your Google Business Profile, Zillow, Realtor.com, LinkedIn and your brokerage roster page. Same facts, different lengths. Retire anything you no longer use rather than leaving it live.

3. The proof. Add structured data so the facts are labeled rather than inferred: person or organization, reviews, listings, service area.

4. The portals and directories. Claim what is unclaimed, correct what is wrong, and kill the duplicates the name check turned up. Most agents find a second phone number here they forgot existed.

5. The reviews and the mentions. Reviews are corroboration. Third-party mentions are how you get into the source pool from check six. Local press, association pages, podcasts, a market report someone else publishes. One real placement beats a month of posting.

6. The site. Last, because it is the biggest and because everything above makes it pay off more. Unless checks five and seven both failed, in which case read the next section first.

That is the order the six signals, in order framework follows on a real estate engagement, and it exists for a reason: an assistant will not repeat a claim it can find in only one place.

If you run marketing for a brokerage

The same seven checks work at the firm level with two changes. Run the short-list check on the brokerage name and on your top three producers. And run the story check across the roster pages rather than one bio. A brokerage with forty agents has forty versions of its own address live somewhere, and that is usually the first thing worth fixing.

When the site is the problem

There is a version of this where the profile work does not help.

If checks five and seven both failed, your website is not a weak signal. It is an absent one. The assistant arrives, reads a shell, and leaves with nothing to confirm the profiles against. You can correct every directory listing you own and still get hedged answers, because the one source that should be authoritative about you has no readable content in it.

The symptom is easy to miss. The site looks fine. It may look expensive. Photography loads, the IDX search works, the contact form submits. None of that is in the HTML. The neighborhoods you serve, the bio you paid someone to write, the listings that prove you work this market: all of it arrives after the page is read, and for a crawler, after is the same as never.

This is a build, not a tweak, and it is the honest answer when an agent asks why their visibility will not move. What a search-ready website build covers walks through what changes.

Where this ends and the Blueprint begins

These seven checks tell you what is broken. There are three things they cannot do.

They cannot tell you how you compare with the four agents who keep getting named instead of you, which would mean running every check on each of them. They cannot measure change over time, because one reading is a snapshot and assistants answer the same question differently from week to week. And they cannot tell you which of those competitors the assistants already trust, or why.

That is what the AI and Search Intelligence Blueprint is for. It is $999, it covers your market and your named competitors with a prioritized roadmap, and the $999 credits back if you start a program within fourteen days. It is a different thing from this checklist, and this checklist is a real thing on its own. Run it first.

Frequently asked questions

How do I check if AI recommends me as a real estate agent?

Ask ChatGPT, Gemini and Google AI Mode who the best agents are in your market and specialty, in a fresh chat, with your name left out of the question. If none of the three names you, you are not in the verified set for that query. Repeat a day later, because answers move.

Why does ChatGPT get my brokerage or specialty wrong?

Because two of your public profiles disagree and it picked one. Assistants read your site, your Google Business Profile, Zillow, Realtor.com and LinkedIn, then state what most sources support. An old brokerage still live on a forgotten profile is the most common cause.

Can I get an AI visibility audit for my real estate website?

Yes, and you can run the first version yourself with the seven checks above in about twenty minutes. A paid audit adds competitor comparison, source-level detail and measurement over time.

How often should I re-run these checks?

Quarterly is enough for most agents, plus any time something changes: a brokerage move, a new designation, a new market, a new team name. Re-run the referral check the week after any change, because an out-of-date profile is what the assistant will find first.

Does this replace SEO for realtors?

No. It is the diagnostic layer on top of it. The checks find which signal is breaking and the search work repairs it. Most of these fixes are the same ones that make you findable in Google, because both readers want a record that agrees with itself.

Where to start

If you ran all seven, you have something most agents do not: a specific list of what is wrong, in the order it needs fixing.

If you want the fixes written out, the Luxury Agent's AI Visibility Checklist is free and walks each signal with what a pass looks like and what to do when it fails.

Related reading: AI search strategies for luxury agents.

Waste less. Grow smarter.