Knowledge base

Search engines and AI assistants (Google, ChatGPT, Perplexity and others) read a hidden layer of labels on a web page. The labels say this page is about an agent named Maya Castillo, who works for Harbor Point Realty, holds license SL1234567, and serves Cooper City. That layer is called structured data, and the standard for it is schema.org.

If the labels are missing or wrong, the machines guess. If they are right, the machines can say who you are with confidence. That matters most for the real-estate community, where licenses, brokerages, teams and brands are easy to mix up.

This site does three jobs, all free:

There are no accounts, no passwords and no pricing. Nothing you type into Profile leaves your computer unless you press a button that says it will (see §11).

What "passes" means. A pass means the labels are correctly built. It does not mean your license number is real or your phone number is right. The tool checks structure and vocabulary, not facts. Every result says so.

You will see these words in the tool. Each one is explained once here.

1. Structured data / JSON-LD. The labels, written in a format called JSON-LD. You never write it by hand. You copy a block of it and paste it into your page. It is invisible to visitors.

2. Entity. A thing the labels describe: a person, a brokerage, a website, a place, an event. One page's labels usually describe several entities and how they connect.

3. The graph. Those entities, plus the links between them ("Maya works for Harbor Point"). The tool shows this as a small family tree on the results screen.

4. ID. Each entity gets a unique address so other labels can point at it without repeating everything. Your page URL and site URL are what the tool builds IDs from, which is why they are required.

5. Check result levels. Every finding is one of three kinds:

   You see	Meaning	What to do
   ✖ Fix required	The labels are wrong in a way machines will trip on.	Fix it. The page does not pass until you do.
   ▲ Should fix	Valid, but likely to confuse a search engine.	Fix it when you can.
   • Notes	Worth knowing. Not a problem.	Read, then move on.

   The headline reads Passes · 0 to fix or Needs fixes · N to fix. Only "Fix required" items count toward N.

Do this once, in this order.

1. Open /profile and press Start from the sample. The sample is a made-up agent (Maya Castillo, Bayline Home Group). Click Next through all seven steps to see what it asks for, then press Save profile. Nothing is sent anywhere.

2. Open /builder (the Create link in the top bar). The page type is already Agent or team profile. In Page URL, type any web address that starts with https://. The code appears on the right as you type, and the headline should read Passes · 0 to fix.

3. Press Copy JSON-LD. That is the finished code for an agent page.

4. Open /audit (the Check link). Choose Paste code, paste what you copied, press Check. You should see the same pass. Change one word of a property name in the box (for example name to nam) and check again: you now get a Fix required, with the message Did you mean "name"?

If those four steps worked, everything below is variations on them.

The sample profile is fictional on purpose. When you create your own profile, use real details only for people who have agreed to be listed.

The pill-shaped bar at the top of every page has: Check · Create · Profile, and a search button.

• Search (⌘K on Mac, Ctrl+K on Windows). Opens a quick-jump box. It lists page types, actions, a glossary of the technical words, and your ten most recent checks. Type a few letters and press Enter.

• Footer links. Help (this knowledge base), Privacy, Terms, Accessibility, Dataset (the license-authority list, see below).

• A page that doesn't exist shows a "This address isn't on the map" page with a button that opens the search box, and a link back to the home page.

• /bot. A public page describing the fetcher behind Check (§12).

• /licenses ("Dataset"). A free list of the real-estate licensing authorities for every US state and Canadian province, with the license categories each issues. It can be downloaded as JSON or CSV. The Profile wizard uses the same list to prefill your license fields.

The site works on phones. Two-column screens stack into one.

Why bother. Every page you create needs the same facts: who you are, your license, your brokerage. Profile asks once. Create reuses it.

Where it lives. Only in your own web browser. Nothing is uploaded. This has two consequences:

• A different computer or browser has no profiles. Download your profile (below) to move it.

• Clearing "site data" in your browser deletes your profiles. Download a copy of each profile you care about.

The Profile screen

Under Saved profiles, each profile has Edit, Download YAML, Download JSON and Delete.

If a profile was saved under older rules and no longer loads, it is listed with only Download and Delete. Download it so you don't lose it, fix it, and import it again.

If an imported profile has the same ID as one you already have, you are asked whether to overwrite or import as a copy.

The seven steps

A progress bar shows where you are. Back and Next move between steps. cancel leaves without saving.

Required fields are marked with a star. Some fields clean themselves up when you save: phone numbers become international format (+19545550142), web addresses go to lower case, and a licensing authority written in capitals gets its normal spelling.

Brand or brokerage? (the part people get wrong)

Real estate often has three layers:

You (the agent) → your brand or team (the name on your signs) → the brokerage of record (the licensed company).

The tool handles two common cases:

• A real team. You work for the team. The team is a real organization under the brokerage.

• A solo brand. Your brand is just your marketing name. You still work for the brokerage; the brand rides along beside you.

The wizard decides which one applies from what you enter, and the diagram on the last step shows the result. Licenses go where they belong: the brokerage's license on the brokerage, your license on you, and the brand carries none.

Open Create (/builder). The form is on the left. The finished code is on the right and updates as you type (after a short pause). The headline above the code shows pass or fix.

If you have no profile yet, Create says so and links to Profile. If you have several, a Profile dropdown appears.

Pick the page type

A property-listing type does not exist yet. Listing pages are the hardest case (MLS and IDX feeds, co-listing). Existing listing code can still be checked.

Help article is meant for software, coaching and service companies rather than agents. It needs a provider profile: in Profile, choose New profile, then the path for a company that serves real estate professionals. The author (a named byline) is optional for a provider.

Open house dates. Pick the start and end with the date and time pickers. The tool works out the time zone from what you pick, so you do not type any time-zone codes.

Finding Wikidata links. A Wikidata link tells machines exactly which place you mean, so "Miami" can't be mistaken for another Miami. The tool can look them up for you; nothing is looked up until you press a button, and nothing is filled in until you pick a match.

• Service areas (Profile, step 4). Type the place, then press Find on Wikidata under it. It lists the states, counties and cities with that name in your profile's state or province. Pick one to fill the link.

• Neighborhood guide (Create). Type the place's name and the larger area it is part of, then press Find both on Wikidata once. It finds the larger area in your state, then the place inside it, and lists the pairs, for example Coconut Grove, part of Miami. One pick fills both links. If the larger area has already been linked, it uses that link. If Wikidata publishes a map point for the place and the latitude and longitude boxes are empty, the pick fills them too, rounded to the precision Wikidata gives. Numbers you typed are never replaced.

• When nothing is found. Many small neighborhoods aren't on Wikidata, or aren't recorded as being inside their city. Leave the link empty, or paste one you found yourself. If only the larger area is found, you can still pick it on its own.

• It needs your state. The buttons only work once your profile has a state or province (Profile, step 1).

The important question: "Does your site already add structured data?"

Most website platforms and SEO plugins already add some labels. If you add a second, competing set, you create duplicates. So the form asks:

• No, this is the only markup. The code stands alone. Use this if you are not sure and a Check of your page shows no existing labels.

• Yes, link to what is already there. The code attaches to what your platform already outputs. Two extra boxes appear: Your page's existing ID and Your site's existing organization ID. A developer or your plugin's settings can tell you these. They look like https://yoursite.com/page/#webpage and https://yoursite.com/#organization.

When unsure, run Check on your live page first (§8).

Reading the right-hand side

1. A list of entity types across the top (agent, person, organization and so on).

2. "N nodes" and a family tree of what was built. Click an item to highlight any finding about it. The other way round, press Show on a finding to highlight its item (Hide clears it).

3. Passes · 0 to fix or Needs fixes, then any Notes. Notes about EducationalOccupationalCredential being "provisional" are normal and not a problem: schema.org publishes that label at a staging address while it finishes standardizing it.

4. Copy code for your site, with Copy JSON-LD and Download .jsonld buttons, then the code itself.

If a required field is empty or wrong, the right-hand side shows a short plain message in a red box naming the field, for example pageUrl: Expected an http(s) URL. Fix the field and the code returns.

The tool never publishes anything. Create only produces text for you to copy.

Create gives you a block that begins {"@context": "https://schema.org", .... It must be wrapped in a script tag and placed in the page. The wrapper is:

<script type="application/ld+json">
  ...the code you copied...
</script>

Where it goes depends on your platform. Typical options:

• WordPress. An SEO plugin with a "custom schema" or "header/footer scripts" setting, or a "Custom HTML" block on that one page.

• Page builders and site platforms. A "custom code", "embed" or "head code" field on the individual page. Use the per-page field, not the sitewide one, since each page's code names its own URL.

• A developer. Send them the copied code and the page's address.

Rules of thumb

• One page, one set of code: put each page's code only on that page.

• Do not edit the code by hand. Change the form, copy again.

• If the site already outputs labels, use the "link to what is already there" option so you don't get duplicates.

After you publish, always test. Open /audit, choose Enter a URL, paste the page's address and press Check. You should see your entities under What we found and a pass. If something is added by JavaScript after the page loads (common with IDX search tools), also use the live-page check in §8.

Open Check (/audit). It works on your pages and on anyone else's, so it is also useful for studying a competitor's markup.

Two ways to check

Your last ten URL checks are remembered in your browser (and shown in the search box) so you can repeat them. They are not stored on the server.

What happens during a URL check

The site's fetcher reads the site's robots.txt first, then downloads the one page you named, following at most four redirects. It then looks for labels in the page's HTML. The fetch gives up after about 12 seconds. It will not fetch internal or private addresses such as 127.0.0.1, and only uses the normal web ports.

If it cannot check the page, you get a plain explanation and the result is not an error in the site. Common reasons:

When the page could not be fetched, there is no pass or fix verdict and the count reads 0 blocks of structured data. That only means nothing was read, not that the page has no labels. The reason is in the highlighted line underneath. It does not mean your labels are wrong.

The live-page check (for IDX widgets and tag managers)

Some tools (Showcase IDX, IDX Broker, iHomefinder, tag managers) add labels after the page loads, using JavaScript. A normal check reads only the original HTML and cannot see those labels. Some search crawlers cannot either. The live-page check shows you the difference.

One-time setup. After running a URL check, a panel titled Check schema added after the page loads appears. Drag RE Schema Inspector to your browser's bookmarks bar, or press Copy bookmarklet URL and paste it into a new bookmark.

Each time.

1. Open the page you want to test in a new tab.

2. Click the RE Schema Inspector bookmark. A Check tab opens with the page's labels.

3. Or, to keep results in your current Check tab: press Open page and listen first, open the page it gives you, then click the bookmark. The result is handed back and the extra tab closes.

You then see each block tagged Added after load or In HTML source, and a Should fix warning for every type that exists only after the page loads.

Only use the bookmark on pages that are fine to inspect. Do not use it on signed-in or private pages (email, banking, dashboards): it reads the page's address and labels. Details in §11.

Results appear in this order:

1. What we found. The family tree of entities. Click one to jump to its findings. Each shows a small count of its findings.

2. A cross-block warning, if one set of labels points at something defined in a different <script> block on the same page. It is valid, but AI crawlers have not said they combine blocks the way Google and Bing do. If you see this, ask your developer to merge the blocks.

3. The verdict and the three groups (Fix required, Should fix, Notes). Under the verdict is the reminder that it checks structure, not truth.

4. Raw code. The labels exactly as the tool read them, folded away by default.

Each finding has four parts:

What is wrong · Did you mean… (a suggestion, when there is one) · Where (a short path like @graph[0].nam, meaning "the first entity, the property called nam") · Which check found it (for example property-existence).

The "where" path is mostly for developers. You can ignore it and use the plain sentence.

Copying findings. The text of a finding can be selected by hand. Each finding also has a Copy button (it copies all four parts plus the entity it is about), and Copy all findings above the groups copies the whole report, ready to send to a developer.

The most common thing you will see is a misspelled property name. The suggestion usually names the fix.

"Passes" is not the finish line. It means the code is built correctly. Also make sure the facts are right, the labels are on the live page, and a Check of the live URL passes.

There are 26. This is the full list, so you can look up anything you meet. Names appear under each finding.

Is the vocabulary right? (usually "Fix required")

Do the links between entities hold together?

Is it sensible for real estate? (mostly "Should fix" or "Notes")

Only on live pages

The site has no accounts, no cookies, no tracking and no analytics.

To delete everything. Clear this site's data in your browser's settings. That erases profiles and recent checks. The server holds nothing to delete.

The Privacy page describes what the code does. It also has a disclaimer of warranties and a Hold Harmless section that applies to anyone who uses the site, its API or what it generates. This manual is not legal advice. Have your lawyer read the Privacy and Terms pages before you rely on them.

When you check a URL, the fetch comes from a robot named RealEstateSchemaBot. The /bot page explains it publicly. This is what a site owner needs to know:

• It reads robots.txt first and obeys any group that names it.

• It fetches only the page it was asked to check, plus the robots file.

• It does not crawl a site.

To block it from your site, ask your developer to add this to robots.txt:

User-agent: RealEstateSchemaBot
Disallow: /

Checks of your pages will then say they were blocked. You can still use Paste code.

You can skip this section unless you want an AI assistant or another program to use the engine. The site's own Create and Check screens are built on the same interface, so anything they do, a program can do.

Why it exists

Search engines are no longer the only readers. AI assistants look things up, and the people who build them (and the people who build tools for agents) need a dependable way to produce and test labels without clicking through a form. The interface returns the same results as the screens, in a form a program can read.

Where to find it

How it behaves

• One error shape. A failure always comes back as an error with a code and a message that says what to fix. When the request itself is wrong, every problem is listed at once, each naming the part and field.

• Nothing is silently ignored. Unknown or misspelled fields are rejected and named.

• Engine version. Every answer carries the engine version (currently 0.4.0). If it changes, the same input can give a different result, so a program that stores results should store the version too.

• Limits. Requests over 1 MB are refused. There are no API keys. Abuse is handled by address: the host allows up to 120 requests a minute per network address, and checking a URL has its own smaller limit, with a waiting period that grows for repeat offenders. A limit answer says how long to wait. Wait the full time, then try again.

• Servers and scripts only. Other websites' pages in a browser are not allowed to call the interface. Programs, AI tools and command-line tools are.

• Privacy is the same as the screens. Nothing sent is stored. See §11.

Local AI tool (MCP)

MCP is a standard way for an AI assistant to use outside tools. A developer can run the engine as a local tool for an assistant. It offers five read-only tools:

Checking code automatically

A developer can also run the same checker in an automated pipeline, so a change that breaks a page's labels is caught before it goes live. Ask your developer for the "CI action".

None of this is needed to use the site.

Terms beginning with a capital letter or in camelCase are schema.org names.

@id

The unique web address that tells search engines exactly which property, agent, or office a piece of data refers to.

about

The main subject of a page or event, such as a specific property listing or a neighborhood guide.

AEO

Answer Engine Optimization: being the source AI answer engines cite.

Apartment

A condo, apartment, or attached residence.

Bookmarklet

A bookmark that runs a small tool on the page you are viewing. Here: RE Schema Inspector.

brokerage of record

The licensed company that legally holds your real estate license.

Crawler

A robot that reads web pages for a search engine or AI.

domain/range

The rules for where a specific detail can be used. For instance, 'number of bedrooms' can be used on a house, but not on a person.

E-E-A-T

Google's Experience, Expertise, Authoritativeness, Trust framework for content quality.

E.164

The international phone number format search engines expect, e.g. +19545550142.

EducationalOccupationalCredential

The proper way to format a real estate license or professional certification (like a CRS or GRI) for search engines.

Engine version

The release number of the code that builds and checks labels. Currently 0.4.0.

Entity

A person, company, site, place or event described by the labels.

enumeration member

A specific, pre-approved option you can choose for a field. For example, an open house status must be exactly 'EventScheduled' or 'EventCancelled'.

GEO

Generative Engine Optimization: the same idea, for generative AI search.

Graph

The entities plus the connections between them.

House

A general category for a home or residence.

ID

A unique address given to an entity so other labels can refer to it.

IDX

The feeds that show MLS listings on an agent's site.

ItemList

A formatted list of items. We use this to organize comparison points, like the pros and cons of two different neighborhoods.

JSON-LD

The format the labels are written in.

knowsAbout

The specific neighborhoods, property types, or real estate topics an agent has proven expertise in.

Lint / linter

Automatic checking, like a spell-checker for structured data.

localStorage

Your browser's own storage, where profiles saved here never leave your device.

MCP

A standard way for AI assistants to use outside tools.

Node

One entity in the graph. "19 nodes" means 19 entities.

Offer

The specific terms of a sale or rental, including the price and availability status.

Person.brand

A real estate agent's personal brand name or team name, kept distinct from their official brokerage.

pinned release

The exact schema.org version this tool checks against, so results don't shift under you.

Place

A physical location, like a neighborhood, city, or open house address.

plugin stitching

Connecting your custom real estate data to the basic code your website platform already creates, rather than creating messy duplicates.

PostalAddress

The structured format for a property's mailing address.

Profile

Your saved identity details (license, brokerage, areas, links).

Rate limit

A cap on requests per minute that protects the site from overload.

RealEstateAgent

The official type used by search engines to recognize a real estate agent, team, or brokerage.

RealEstateListing

The standard format for a property that is currently for sale or rent.

Recipe

One page type's template (agent, article, neighborhood, FAQ, open house, help article).

robots.txt

A file on a website telling robots what they may fetch.

sameAs

A link to your verified profiles elsewhere online, like your Zillow, Realtor.com, or LinkedIn page.

schema.org

The shared vocabulary of labels that search engines and AI understand.

SingleFamilyResidence

A standalone, detached home or property.

Structured data

The hidden labels on a page. Same as JSON-LD for our purposes.

Vocabulary

The list of labels schema.org allows.

Wikidata / Wikipedia link

A link that tells machines exactly which place or thing you mean.

worksFor

The real estate brokerage or team that an agent is licensed with.