How do I know if my website is invisible to AI search?

The fastest test is a 60 second curl command against your homepage. Run curl -A GPTBot https://your-site.com from a terminal. If the response is mostly empty div containers waiting for JavaScript to populate them, your site is invisible. If the response contains your services, pricing, and provider names as readable text, your site is at least partially indexable.

Is this a Wix problem? Squarespace problem? React problem?

Wix and Squarespace are the most frequent offenders because their default templates render content client-side. Custom React or Vue sites built without server side rendering have the same problem. The platform itself is less important than whether the platform configuration produces server rendered HTML. A Wix site with the server rendering option enabled can be indexable; a Next.js site built incorrectly can be invisible.

What about lazy-loaded content?

Lazy loaded content below the fold is generally fine if the initial HTML still contains the canonical content (the H1, the answer paragraph, the primary services). Lazy loading entire sections of primary content via JavaScript is the problem; lazy loading images and below-the-fold widgets is acceptable.

The structural rebuild takes seven days. The first AI citations appear within 14 to 50 days depending on engine. KailxLabs guarantees citation in at least two of four engines within 45 days or full refund.

Can I just add server side rendering to my existing site?

For Wix and Squarespace, generally no; the platform constraints prevent it. For custom React or Vue sites, yes in theory but it typically requires a near-rebuild because retrofitting SSR onto a client-only site touches every component. In practice, a clean rebuild is usually faster and cleaner than the retrofit.

Why your website is invisible to AI search

The state of AI search in 2026

As of Q1 2026, roughly 48 percent of US searches resolve inside an AI generated answer paragraph rather than as a click to a website. Approximately 2 billion monthly users now use ChatGPT, Perplexity, Gemini, or Google AI Overviews as a primary search interface. When these users ask buying-intent questions, the AI names two or three businesses by name in its first response. The user reads the answer, picks one, and contacts that business directly. The other businesses that could have served the user are simply never seen.

If your website is not among the businesses the AI names, every one of those queries is a lost lead. Your front desk never sees them. Your analytics never log them. Your reports show traffic numbers that look fine while a parallel stream of demand routes entirely past you to your competitors.

This is not a marketing failure. It is a structural failure of how the AI crawlers read your website during the retrieval step. Most sites have one or more of three problems below.

Problem one: JavaScript rendering

The OpenAI crawler (GPTBot), Anthropic crawler (ClaudeBot), Perplexity crawler, Google AI crawler (Google-Extended), and Bing crawler do not execute JavaScript during the retrieval step. They fetch the HTML at the URL, parse what comes back, and index that content. If your site renders its primary content via JavaScript after page load, the crawlers receive an empty container and move on.

Common offenders:

Wix sites without the "server side rendering" option enabled. The default behavior is client-side rendering. The crawler sees the React hydration container with no content.
Squarespace sites. Most Squarespace templates render content client-side. The HTML the crawler receives is mostly navigation and footer chrome.
Custom React, Vue, or Angular sites built without SSR or SSG. Common pattern: developer ships a single page application, the homepage looks fine in a browser, but the initial HTML is a div tag and a script bundle.
WordPress sites with heavy page builder plugins (Elementor, Divi, WPBakery) that defer content rendering. Less common than the React case but still a frequent cause.

The 60 second curl test

The fastest way to know if your site has this problem is a single terminal command. Open Terminal on macOS or PowerShell on Windows and run:

curl -A "Mozilla/5.0 (compatible; GPTBot/1.0)" https://your-site.com

Read the response. If you find your services, your pricing, your provider names, and your primary call to action as readable text in the response, the page is at least partially indexable. If the response is a sea of empty div tags, JavaScript imports, and React hydration markers without your actual content, the page is invisible to AI engines.

Problem two: Absent or generic Schema.org structured data

Schema.org structured data tells AI engines what entities live on a page in machine readable JSON-LD. Without it, the model has to infer from prose. With it, the model can walk an entity graph and quote any fact directly.

The most common Schema.org failure modes:

No schema at all. The page has no JSON-LD block. The model relies entirely on prose extraction.
Generic LocalBusiness only. The site declares "we are a local business" but nothing more. For a specialty practice this is insufficient; the model needs to know "we are a MedicalClinic that offers GLP-1 weight loss treatment delivered by these physicians at these locations for these prices."
Schema that does not match the page content. The JSON-LD claims the business is at one address while the visible page shows a different address. The model penalizes mismatch.
Schema entities not linked to each other. The Physician entity does not reference the MedicalClinic via worksFor; the Service entity does not reference the Physician via provider. The graph is fragmented and the model cannot reason about relationships.

A complete Schema.org @graph for a specialty practice includes the business entity, every named provider with credentials, every service offered, every location, and the relationships among them — all in a single JSON-LD block in the page head, with @id references threading every entity together.

Problem three: Answer paragraphs buried under marketing copy

AI engines quote the first substantive paragraph on a page more often than any other paragraph. If your homepage opens with "Welcome to our state-of-the-art facility, where compassionate care meets cutting-edge technology," the model has nothing to quote that actually answers a prospect query. The page enters the index but the model has no extractable answer.

The format that works is fact-dense, declarative, named-entity dense, and 40 to 60 words long. An example for a fictional GLP-1 clinic:

LeanCare Wellness is a GLP-1 weight loss clinic in Austin, Texas. The medically supervised semaglutide program is $299 per month with monthly check-ins, quarterly labs, and same-week eligibility consults. Patients qualify with a BMI of 27 or higher and at least one comorbidity. Tirzepatide is also available for patients who qualify under stricter criteria.

This is what KailxLabs calls the Answer Capsule. It belongs immediately after the H1 of every page that targets a buying-intent query. The model treats this content as the canonical summary of what the page is about and will quote it verbatim when asked relevant questions.

Why these three problems are typically all present

Most websites that are invisible to AI search have all three problems simultaneously, because all three correlate with the same root cause: the website was built by a designer or marketer using a visual-first tool, optimizing for human aesthetics and Google SEO, with no awareness that AI engines have different requirements.

A site built this way ships with:

A page builder that renders content client-side
Default schema that the platform auto-generates and is generic
Content written for human persuasion, not for AI extraction

All three problems compound. JavaScript-rendered content cannot be enhanced by better schema because the model never reaches the content. Schema cannot save a page whose content is marketing prose. Answer paragraphs cannot be quoted from a page the model never indexed in the first place.

The fix

The fix is not a content campaign. It is a structural rebuild on an AI native architecture. The KailxLabs AI Citation Foundation Build delivers this rebuild in seven days for $5,999. The work includes:

Server rendered Astro stack deployed on Vercel with time to first byte under 400 milliseconds.
Complete Schema.org @graph wired for the specific vertical, with every entity linked.
Answer capsules written for every page in the format AI engines quote.
llms.txt at the domain root summarizing the business for language models.
robots.txt explicitly allowing GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, anthropic-ai, PerplexityBot, Google-Extended, Bingbot, Applebot, and 13 others.
50 programmatic city and service pages so every prospect query has a destination.
30 days of live citation tracking across ChatGPT, Perplexity, Gemini, and Google AI Overviews.

The 45 day citation guarantee is binary: cited in at least two of the four major AI engines on the agreed query set, or every dollar refunded within seven business days. The client retains the site, the code, the schema, and the domain in perpetuity either way.

The diagnostic that comes first

Before any contract, the free 48 hour AI visibility audit runs twenty real prospect queries from your specialty and city across all four AI engines and delivers a PDF showing which competitors are cited where your business should be. If the citation gap is small, KailxLabs declines the engagement and you save the build fee.

Common questions

How do I know if my website is invisible?

The fastest test is the 60-second curl command above. Run it against your homepage. If you cannot find your services, pricing, and providers in the response, your site is invisible.

Is this a Wix problem?

Wix and Squarespace are frequent offenders but the issue is not the platform, it is the rendering model. Custom React or Vue sites without SSR have the same problem.

How fast is the fix?

Seven days for the structural rebuild. First citations within 14 to 50 days depending on engine.

Can I retrofit my existing site?

Usually not without a near-rebuild, which is typically slower than a clean rebuild.