Open your own website on your phone right now. Not the pretty version you show clients. Just load it.
Now imagine stripping out every image, every animation, every piece of JavaScript, and reading only what is left. Whatever text survives that stripping is what ChatGPT sees when it decides whether to recommend you.
For a lot of real estate websites, what survives is almost nothing.
Part One Told You to Publish. This Is the Part Underneath It.
In the first article on AI visibility for real estate agents, we covered the Local Falcon study: 37,500 AI searches across the 100 largest U.S. cities, and 91.5 percent of ranked, website-publishing agents never cited once.
The conclusion was to publish credible local authority content so AI systems have something to work with.
That advice still holds. But it assumes something nobody checked first.
It assumes the machine can actually open your page and read the words on it. For most real estate websites, that assumption is wrong.
I Went Looking for a Tool. I Did Not Find One I Trusted.
After that first article went out, the question I got back most was the obvious one. How do I know if this is me.
Fair question. So I went looking for something I could hand people.
What I found was disappointing in three predictable ways. The serious technical crawlers are built and priced for enterprise SEO teams, and they answer questions an agent is not asking. The free ones mostly wanted an email address before they would show a number, which tells you what the product actually is. And a lot of them were still scoring for 2019 Google SEO, measuring keyword density and backlinks while ignoring the thing that actually decides this now: whether an AI crawler can fetch the page and find words on it.
Not one of them led with fetchability. Several never checked it at all.
So I built the thing I wanted to hand people. That is the only reason it exists.
It is free, there is no signup, and the entire scoring model is published on the page so you can argue with it. What follows is what I learned building it, and why the weighting looks the way it does.
AI Crawlers Do Not Run JavaScript
This is the technical fact that quietly decides who gets cited, and almost nobody in real estate is talking about it.
When a search engine like Google crawls your site, it can execute JavaScript. It waits for your page to build itself, then reads the finished result. That is why a modern IDX website full of dynamic content still ranks.
AI crawlers do not work that way. GPTBot, OAI-SearchBot, ClaudeBot, and PerplexityBot request your page and read the raw HTML that comes back. Before any JavaScript runs. They do not wait. They do not retry.
A server log study by Vercel and MERJ found no evidence of JavaScript execution across major AI crawlers. In an analysis of more than 500 million GPTBot fetches, researchers found zero JavaScript execution. GPTBot downloads JavaScript files roughly 11.5 percent of the time and never runs them. ClaudeBot downloads them about 23.8 percent of the time and never runs them either.
So if your website is built as a JavaScript shell, which describes a large share of agent sites and nearly every template-driven IDX platform, here is what OAI-SearchBot receives: an empty container and a pile of scripts it will not execute.
You are not losing that referral to a better agent. As far as the engine is concerned, you were never a candidate.
The Gemini Exception, and What It Explains
There is one major exception, and it retroactively explains something from the first article.
Google Gemini runs through Googlebot's rendering service. It can execute JavaScript. It sees the finished page the way a human does.
Now go back to the Local Falcon finding: Gemini pulled more citations from agents' own websites than any other platform studied.
That was reported as a platform preference. It is probably not a preference at all.
Gemini cites agent websites more often because Gemini is the only engine that can read them. The other platforms are not ignoring your site. They are receiving a blank page.
That reframes the whole problem. The gap between how you perform on Gemini and how you perform on ChatGPT is a rendering gap, and it is fixable.
Why Fetchability Outweighs Everything Else
When I built the free AEO Site Checker at aeo.jmanai.com, fetchability got the heaviest weight in the entire score. Forty points out of a hundred, more than every content signal combined.
That is deliberate, and the reasoning is simple.
Most scoring factors are a matter of degree. A thin meta description is worse than a good one, but it is not fatal. Fetchability is not a matter of degree. It is binary.
A page a bot cannot fetch gets cited exactly as often as a page with nothing on it. Zero times.
You can write the most rigorous, well-sourced local authority page in your market, follow every rule from the first article, cite your criteria, name other agents, date the page, and if your host throws a bot challenge at OAI-SearchBot, none of that work exists.
Fix the fetch first. Everything else is downstream of it.
What the Checker Actually Looks At
Paste any page, get a score from 0 to 100 in about ten seconds. No signup. Six categories, weighted by how much each one can remove you from an answer.
Fetchability, 40 points. Whether a bot can fetch the page, what your robots.txt says to the nine crawlers that actually produce citations, whether your text is server-rendered, HTTPS, page size.
Answer engine signals, 25 points. JSON-LD entity markup, extractable body copy, llms.txt, site depth, author and sameAs provenance, dateModified.
Core SEO, 21 points. Title, meta description, sitemap, canonical, Open Graph, Twitter card, language. Google AI Mode is still search underneath, and the first article showed it pulls 72.5 percent of its citations from Google Business Profile.
Semantic HTML, 13 points. Landmarks, one H1, a clean heading outline with no skipped levels, alt text. This is the skeleton a parser uses to locate the answer inside your page.
Content quality, 6 points. Answer-first opening paragraph, real numbers, question headings, quotations. These are weighted lightest on purpose because they are proxies rather than measurements. They are grounded in the Princeton GEO research, published at ACM SIGKDD, which tested roughly 10,000 queries and found that citations, quotations, and statistics measurably improved visibility in generated answers.
Contact signals, 5 points. Phone, address, hours, service area. Only scored for businesses and people, so a blog post is not penalized for lacking them.
A is 90 and up. B is 80. C is 70. D is 60. Below that is an F.
How to Read Your Score Without Panicking
A low score is not a verdict on your business. It is usually a verdict on a platform decision someone else made for you.
Most agents did not choose their website architecture. A brokerage, a vendor, or a template did. If you score poorly on fetchability, that is a vendor conversation, not a personal failing.
Work the fixes in the order that recovers the most points, which the report gives you directly. Broadly, that order is:
First, get the page fetchable and server-rendered. If your text only exists after JavaScript runs, nothing else on this list matters. Ask your provider one question: is my content in the initial HTML response.
Second, check robots.txt. Plenty of agents are blocking AI crawlers without knowing it, sometimes because a security plugin or host default did it for them.
Third, add structured data and a real bio in real paragraphs. Not a graphic with text baked into it. Words a parser can lift.
Fourth, tighten the SEO and semantic basics.
Then re-run and see what moved.
What the Score Cannot Tell You
Being honest about the limits, because the first article held other people's tools to this standard and mine should not get a pass.
The score is a heuristic, not a guarantee. It measures whether an engine can technically fetch, read, and extract your page. It does not measure whether an engine will choose you.
Citation decisions also depend on your Google Business Profile, your portal presence, reviews, links, and factors nobody outside those companies can see. A perfect score does not put you in the answer. A failing score keeps you out of it. Those are different claims and I am only making the second one.
It also checks one page at a time, not your whole site, and it cannot see what the engine does after the fetch.
My Tool Is a Funnel Too. Use It Anyway.
In the first article I wrote about a free visibility tool from another company, and the line was: use the free tool, know why it is free.
I am not going to dodge my own standard just because the tool is mine now.
That standard applies to me.
The AEO Site Checker is free and has no signup and no email gate, and it is also marketing. I train brokerages and agents on this material. I build AI systems for real estate businesses. A tool that shows you a problem I happen to solve is not a neutral public service, and I am not going to pretend otherwise.
Here is why I think you should run it regardless, and it goes back to why it got built in the first place.
I did not build this to generate leads. I built it because I kept getting asked a question I had no good answer for, and every existing answer was either too expensive, gated behind an email, or measuring the wrong decade. The business case came second. You can believe that or not, which is exactly why I published the scoring model instead of asking you to take my word for it.
The scoring logic is published on the page. Every category, every weight, and the reasoning behind why fetchability carries 40 points and content quality carries 6. You can disagree with the weighting. You can take the report to any developer and have them verify it independently, because these are objective technical checks, not opinions.
A tool that tells you its own scoring model is a tool you can argue with. That is the standard I would want applied to me.
The Score Is a Starting Line, Not a Grade
Here is how I want you to use the number.
It is a baseline. A snapshot of where you are standing today, before you have done anything. Nobody is handing out report cards, and a 41 is not a judgment on your career. It is a measurement of a starting position, and you cannot improve a position you have never measured.
What matters more than the score is the ordering underneath it. The report does not just tell you what is broken. It ranks the fixes by how many points each one recovers, which is another way of saying it ranks them by impact.
That ordering is the actual product. Most agents do not have an AI visibility problem so much as a sequencing problem: real effort spent on the thing that moves the needle least.
Adding structured data to a page that OAI-SearchBot cannot fetch is busywork. Rewriting your bio for a page that renders only in JavaScript is busywork. Both feel productive. Neither changes anything. The report exists to stop you from spending a quarter on the wrong half of the problem.
Why This Is a Move-Now Problem
I am not going to manufacture panic. But I will tell you plainly why I would not sit on this.
Buyers and sellers are already asking these engines who to call. That behavior is here, not coming. Every month your site is unreadable is a month the engines are answering questions about your market with somebody else's name in the answer.
The second reason is more practical. Most of what the report surfaces is a one-time technical fix, not an ongoing expense. Getting your content server-rendered, correcting a robots.txt file that is blocking crawlers you never meant to block, adding entity markup. You do that work once. The visibility it buys keeps paying after the invoice is closed.
This is the rare problem where the fix is cheaper than the delay.
And the third reason is that most agents in your market have not run this check yet. That window does not stay open indefinitely.
What to Do This Week
Run your own site through the checker. Use your actual bio or about page, not just the homepage, since that is the page an engine needs in order to describe who you are.
Run one competitor who does show up in AI answers. The gap between the two reports is your roadmap, and it is usually more instructive than your own score in isolation.
Send the fetchability section to whoever runs your website. If your text is not in the raw HTML, that is the entire conversation, and it is a vendor question rather than a marketing question.
Then go back to the first article and do the other half. Fix the Google Business Profile. Draft the honest local authority page. Publishing and being readable are two separate jobs and you need both.
Then Get a Second Set of Eyes on It
The report will tell you what is wrong and what order to fix it in. What it cannot do is tell you what any of it is worth inside your specific business.
That part takes a conversation. Whether your problem is your website platform or your Google Business Profile or the fact that you have never published anything a machine could cite. Whether this is a two-hour fix or a two-month rebuild. Whether it is even your highest-leverage AI project right now, because for some agents it honestly is not.
Bring your report to a call and leave with a plan. That is the whole offer.
Book time with me here: https://jmanai.com/book-discovery-call
And if you would rather work with someone else, do that. Take the report to any AI strategist you trust, or to your brokerage's tech team. The report is yours and it is readable by anyone. What I care about is that somebody who understands this looks at your numbers with you, rather than the report sitting in a tab until you forget you ran it.
The agents who lose here are not the ones who score badly. They are the ones who never check.
Want Help Getting Found?
Reading about AI visibility and actually fixing yours are two different sports.
Check your site free: https://aeo.jmanai.com/
Talk it through: book a discovery call and bring your report.
Done for you: I build custom AI chatbots and visibility systems for real estate agents at Big Brain Chatbots.
Learn it yourself: come into AI-Cademy, where we work through this live every week.
Ready to go all in? The 6-week AI-Cademy Bootcamp.
One question before you go: what did your site score, and was the number higher or lower than you expected?
Sources
Vercel and MERJ, AI crawler rendering behavior in server logs.
Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan, Deshpande. GEO: Generative Engine Optimization. ACM SIGKDD, 2024.
The Realtor AI Visibility Index, Local Falcon, August 2026.
Nick Pipitone, Inman, August 2026.
