AI Visibility, Plain English

Your AI visibility is what determines whether ChatGPT, Perplexity and Google's AI Overviews ever mention your business — explained without jargon, and backed by a live audit of 30 real companies.

Written by Stuart Henderson, Principal Consultant · Stuart Henderson Consulting GEO, AEO & structured data for global businesses

Why AI visibility matters now

More people are starting their search for a business, a property, or a service by asking an AI system a question directly, instead of typing into Google and clicking through a list of links. That shift is no longer a prediction — it's measurable. Google searches in the United States ended without a single click 68% of the time in the first four months of 2026, up from 60.45% two years earlier, according to SparkToro's analysis of Similarweb clickstream data. AI Overviews now appear on more than one in five Google searches, and when they do, click-through rates fall by nearly 60%.

A business can be well-established, fully licensed, and highly rated, and still be invisible in these AI answers. Nothing is wrong with the business — its website was simply never set up to communicate with a machine reader. This guide closes that gap, term by term.

How AI "sees" a website

When a person visits a website, they read the words, look at the photos, and understand the layout. AI systems don't experience a page that way. They read the underlying code instead. Unless that code includes clear, labeled information, the AI is left guessing — piecing together meaning the way someone might skim a document in a language they only half know.

Structured data, schema.org, and JSON-LD

"Structured data" is the general name for hidden labels added to a webpage's code. These labels don't change how the page looks to a visitor — they run quietly behind the scenes. They tag things like "this is the business name," "this is the phone number," "this is a customer review," so a computer doesn't have to infer it.

"Schema.org" is the shared rulebook for those labels — an agreed vocabulary that Google, Bing, ChatGPT, Perplexity, and every other major AI and search system recognize. "JSON-LD" is simply the file format most websites use to write those schema.org labels into a page. It's a technical detail — the important part isn't the format itself, it's whether the labels exist at all. A website with none of this is handing every AI system a blank form where a completed one should be.

Without these labels, an AI system reading a real estate site, for example, has no reliable way to confirm what the business does, where it operates, or that a listing is real. Rather than risk stating something it can't verify, it may leave the business out of its answer entirely. That is exactly the AI visibility failure mode we found repeatedly in the audit below.

FAQ schema

Many websites already have a page of frequently asked questions with real, accurate answers written by the business. FAQ schema is a label that tells an AI system exactly which text is the question and which text is the official answer. That lets the AI quote the answer directly and with confidence, instead of summarizing the whole page and risking a wrong detail.

A page can have good FAQ content and still get no benefit from it. That happens if the content was never labeled as an FAQ in the code. Out of the 30 businesses in our audit, only one — a small Georgian law firm — had FAQPage schema in place. We've marked up the FAQ section at the end of this article with that same schema. This guide demonstrates the exact thing it's explaining.

Business-type labels (Organization, LocalBusiness, RealEstateAgent, and similar)

Beyond labeling individual facts, a website can also label what kind of business it is altogether — a real estate agency, a law firm, a medical clinic, a bank. This is often the single highest-value AI visibility fix available. Most businesses already have every fact this label needs — a license number, an address, a phone number, service areas — sitting in plain text on the site already. Adding the label doesn't require new information; it requires marking up what's already there.

Our audit found this exact pattern repeatedly. Businesses generate a generic WebPage or Organization type via their WordPress theme by default, but never take the extra step to declare LegalService, RealEstateAgent, BankOrCreditUnion, or whichever type actually matches what they do.

robots.txt and AI crawlers — the split that actually matters

Every website can publish a small file called robots.txt — a sign posted at the front door, addressed to automated visitors rather than people. The newest visitors reading that sign are AI crawlers, and there are two different kinds, sent for two different reasons. Confusing them is the single most common mistake we see.

The standard approach in 2026 is to block the training crawlers, keeping content out of model training. At the same time, explicitly allow the search and citation crawlers, so you stay eligible to be cited in ChatGPT, Claude, and Perplexity answers. If your developer or a previous agency added a blanket "block all AI bots" rule sometime in 2023 or 2024, that was a common defensive move at the time. It may now be quietly blocking the exact crawlers that would otherwise cite you.

llms.txt

This is a newer, simpler idea. It's a short, plain-text welcome note written specifically for AI systems, kept at a predictable spot on a website (yourdomain.com/llms.txt). It briefly explains what the business does and points to its most important pages, saving the AI the work of piecing that together from scratch. Not every AI system uses it yet, but it costs very little to add and only helps your AI visibility.

Page headlines (H1 tags) and why duplicates are a problem

Every webpage has a designated "main headline," formally called an H1. It's meant to state in one line what that page is about. Search engines and AI systems lean on that single headline as the clearest signal of the page's topic.

When this happens on one or two pages, it's a minor slip. When it happens across nearly every page on a site, it's usually one shared page template producing the same issue everywhere — fixable in one pass rather than page by page.

Sitemaps and status codes

A sitemap is a directory of every page on a website, handed to search engines and AI tools so they know what exists and can check on it. When something checks in on a page, that page responds with a status code — a one-word answer to a knock at the door:

• 200 — "I'm here, come in." The page is live and working normally.
• 404 — "There's nothing here." The page genuinely doesn't exist anymore.
• 0 / no response — Nobody answered the knock at all. Worth double-checking rather than assuming the worst: it can mean the page is truly gone, or that the automated visitor was blocked, rate-limited, or timed out while the page itself is fine.

Treating a "no response" the same as a confirmed "page doesn't exist" can lead to spending time rebuilding pages that were never actually broken.

What we found checking AI visibility across 30 real businesses

Everything above is a concept. Here's what it looks like in practice. Over a few days in August 2026, we ran a live structured-data check against 30 operating businesses across three markets — Argentina, Nicaragua, and Georgia — ten per country, spanning legal, real estate, banking, logistics, tourism, and healthcare.

Roughly one in three had zero structured data at all — no JSON-LD, no microdata, nothing for an AI system to anchor an entity on. Georgia's businesses were meaningfully behind the other two markets on this sample — a 50%+ zero-schema rate against roughly 20% in Argentina.

Market leaders skip schema more often than smaller challengers do

This wasn't a one-off. It repeated across three separate sector pairs, in two different countries:

• Georgian banking — TBC Bank, the country's largest bank by assets, had nothing. Bank of Georgia, its main competitor, had a correctly-typed BankOrCreditUnion schema.
• Georgian legal services — BLC Law Office, the country's largest firm and the only one with a Batumi branch, had nothing. A smaller firm, Nakashidze & Partners, had both LegalService and FAQPage schema — the best result across all 30 companies checked.
• Nicaraguan banking — Banpro Grupo Promerica, one of Nicaragua's three major banks, had nothing. Banco LAFISE Bancentro, a direct competitor, had a correctly-typed schema.

In every one of these pairs, company size and market position told you nothing about AI-readiness. If anything, the pattern ran backwards: the businesses with the most to lose from being invisible were the ones most likely to be invisible.

Two technical issues beyond "has schema / doesn't"

One Georgian real estate site couldn't be loaded at all due to a live SSL certificate error. That's a browser-level warning that keeps both visitors and crawlers away, regardless of what schema the page might otherwise have. Another had a JSON-LD block that was present but structurally malformed — a @type field containing a raw nested array where a single value belongs. That kind of error typically causes structured-data validators to silently ignore the whole block. It can sit there for months looking fine to a human and doing nothing for a machine.

The takeaway isn't that these specific companies are behind. It's that being a large, credible, long-established business in your market is not a substitute for machine-readable proof of it. The only way to know where you stand is to check.

Check your own AI visibility in 60 seconds

This is the same check used for the 30-business audit above. It works on any website, including ones you don't own or manage — it only reads what's already public in the page's code.

Why AI visibility adds up

None of these individually make or break a business. Together, they determine something simple: when someone asks an AI system a question your business could answer, does the AI have enough to work with to name you — accurately, confidently, and instead of a competitor? A site with clear labels, a welcome note for AI visitors, one clean headline per page, and a trustworthy sitemap gives an AI system every reason to cite it. Without any of that, a site gives the AI nothing to go on — no matter how good the underlying business actually is. As the audit above shows, that includes some of the largest, most established businesses in their markets.

Quick-reference glossary

Content & markup terms

Crawlers, files & technical terms

Frequently Asked Questions

Sources & References

• Search Engine Land: Google zero-click searches hit 68% in early 2026: Study — SparkToro / Similarweb clickstream data, published June 9, 2026.
• Original audit data: SHC internal schema audits of 30 businesses across Argentina, Nicaragua, and Georgia, live-checked August 2, 2026 via the console snippet above — re-run before citing again, as individual sites change.
• Crawler training-vs-search classification cross-checked against current third-party robots.txt / AI-crawler guides (checked August 2, 2026).