A website built by an AI tool can rank on Google and be completely invisible in ChatGPT, Claude, and Perplexity at the same time. Those are not the same problem, and most of the people selling AI-built websites to small businesses are not explaining either one.
I have spent over two and a half years tracking every area of AI that touches marketing, content, and how businesses get found. Websites built by AI tools became a category I watched closely, partly because I have built hundreds of websites over the years, and partly because the “AI website in minutes” pitch started showing up in the same circles as the AI voice agent pitch.
Like my previous post about AI voice agents, there are thousands of people on YouTube and other social channels pitching creating and selling AI websites to small business owners as an easy way to make money. I provided actual quotes and links to the videos in the AI voice agent article. However, for this article, if you want to see what is being sold to those looking to make money online fast, search on YouTube: Make money with AI websites.
Like AI voice agents, there is nothing wrong with selling AI websites. The AI website builders and technology is actually incredible. But what I’ve been seeing is a bit concerning with how some people are pitching the “opportunity”. Not everyone selling AI websites is being honest. And not everyone truly understands the disservice they are providing to small business owners.
There are real use cases for AI websites. But those creating the websites in under two minutes and charging $500 a month or thousands one-time, while not educating the business owner about some of the problems, that is what I have a problem with.
With AI today, anyone with zero experience can create a beautiful website with AI. It takes literally two minutes. And as someone who has built hundreds of websites over the years, I’m still using WordPress and Elementor to build most of my sites by hand. Why? You’ll understand below…
But full disclosure, I do use various AI website builders for different purposes. Some to test new ideas, some to keep an eye on where AI website building it at design-wise, and now I use it to brainstorm website design ideas.
When I create something I like in AI, I then duplicate that design and create it in Elementor/Wordpress when I want to ensure I have proper SEO, AIO, AEO, EAAT, etc.
When I started seeing these sites get sold to business owners who had no idea what was actually under the hood, I decided to research it properly before writing anything.
What the research shows is more nuanced than “AI websites are bad.” The real story depends on one technical variable that business owners have no idea they need to check. And there’s also a content quality problem that affects AI-built sites and traditionally built sites equally.
Both are critical and both affect whether your business gets found.
And one of them connects directly to why generic AI output is becoming a liability, not just a convenience issue. This post covers both sides.
There are a few technical things in here, but I do not go in too deep. This is mainly so you, as a business owner, are educated and know the right questions to ask when someone is offering to build you a website with AI.
The One Test That Tells You Almost Everything About Your AI Website
Open any website in a browser. Right-click anywhere on the page and select “View Page Source.”
A new tab opens showing the raw code the server delivered before the browser did anything with it. Scroll through it and look for your actual content: your headline, your first paragraph, your page title, your meta description.
If those words appear in the source code, search engines and AI crawlers can read your site. If what you see instead is mostly a shell of script tags and a single line that looks like <div id="root"></div> with no actual content, your site is client-side rendered, and a significant portion of the internet cannot read it at all.
That test takes about thirty seconds. It is the most useful thing in this post for a business owner evaluating any website, AI-built or otherwise.
What does “View Page Source” reveal about an AI-built website?
View Page Source shows the raw HTML delivered by the server before JavaScript executes in the browser.
For server-rendered or pre-rendered sites, this includes all page content: headlines, body copy, meta tags, and structured data.
For client-side rendered sites, particularly those built with certain AI website builders, the source shows only a script-loading shell with no visible content.
That distinction determines whether Google, AI crawlers, and other search tools can read and index the page. A site that looks complete in a browser can appear entirely empty to a crawler that does not execute JavaScript.
Why Rendering Architecture Is the Dividing Line
AI website builders split into two categories based on how they deliver pages. The first category pre-renders or server-renders pages, meaning the full HTML content is assembled before it reaches the browser. Framer, Wix (since completing its server-side rendering migration in Q4 2025), and Squarespace all fall into this category.
Google can read these sites reliably, and they perform comparably to traditionally built sites in search.
The second category generates client-side rendered single-page applications, where the browser receives a shell and JavaScript assembles the content after load.
This is critical to know if someone used AI to build your website before May 2026: Until May 2026, Lovable built exclusively in this architecture. Tools like Bolt.new, v0, and Replit generate the same kind of output.
Lovable’s own platform shift tells the story clearly. On May 13, 2026, the company launched what it called a “Discoverability” suite, migrating new projects to server-side rendering and adding automatic pre-rendering for legacy apps.
Their announcement noted that over 200,000 projects ship on Lovable every day. Every project built before that date shipped as a client-side rendered application, meaning a shell that crawlers could not read.
That was not a fringe edge case. It was the default output for one of the most heavily marketed AI website builders in the market.
Which AI website builders are safe for SEO in 2026?
As of mid-2026, Framer, Wix, and Squarespace pre-render or server-render content and are treated by Google as legitimate platforms.
Framer pre-renders every page to static HTML at deployment.
Wix completed a full server-side rendering migration in Q4 2025.
Squarespace has been template and server-rendered throughout.
Durable uses Next.js with server-side rendering and performs well on desktop PageSpeed scores, though it offers limited technical SEO controls and no code export.
Lovable shifted to server-side rendering for new projects as of May 13, 2026; projects created before that date are client-side rendered and require pre-rendering to be discoverable.
Pure SPA outputs from Bolt.new, v0, and Replit are not built for public discoverability.
Google Can Render JavaScript. AI Search Cannot.
This is the split that makes the situation genuinely complicated for business owners, and the part that those selling AI-built websites are probably not discussing.
Google’s crawler renders JavaScript. It operates a headless Chromium rendering engine that processes client-side rendered pages, eventually.
There is a queue delay: client-side rendered content takes hours to weeks longer to index than static HTML, and low-authority new sites wait longer in that queue. But Google gets there.
A Lovable site built before May 2026 will likely appear in Google search results if it has been live long enough. That is true.
What is also true is that GPTBot, ClaudeBot, and PerplexityBot fetch raw HTML only and do not execute JavaScript at all.
Vercel’s analysis of AI crawler behavior across its network found that ChatGPT, Claude, Anthropic, Perplexity, Meta, and ByteDance crawlers all skip JavaScript execution entirely. A Vercel report on crawler behavior documented OpenAI’s GPTBot generating 569 million requests and Anthropic’s ClaudeBot generating 370 million requests across the network in a single month, none of which involved JavaScript rendering.
Glenn Gabe’s case study of a client-side rendered site found that when asked to read specific URLs, ChatGPT repeatedly reported it could not access the content because the page relied on JavaScript rendering, while server-rendered control pages were read and cited accurately.
A client-side rendered AI-built site can rank in Google search and simultaneously produce zero results when someone uses ChatGPT or Perplexity to research businesses in your category.
Given that Adobe Analytics’ Q1 2026 report tracking over one trillion US retail visits found AI-referred traffic up 393 percent year-over-year and converting at 42 percent better than the prior year, that is not a theoretical problem.
Can a client-side rendered website appear in AI search results?
Generally no. GPTBot, ClaudeBot, PerplexityBot, and most major AI crawlers fetch raw HTML only and do not execute JavaScript. A client-side rendered site delivers a script-loading shell to these crawlers with no readable content.
As a result, the site cannot be indexed or cited by AI search tools regardless of how well it performs in Google. Google AI Overviews are an exception because they use Google’s own rendering infrastructure, which processes JavaScript.
However, ChatGPT, Claude, and Perplexity draw from their own crawl data, and client-side rendered pages are effectively invisible to all three.
What the Performance Data Shows
Core Web Vitals measure page speed and stability across three metrics: Largest Contentful Paint (LCP, targeting under 2.5 seconds), Interaction to Next Paint (INP, targeting under 200 milliseconds), and Cumulative Layout Shift (CLS, targeting under 0.1). Google uses these as a ranking signal.
The November 2025 HTTP Archive Technology Report found Wix sites passing Core Web Vitals at 74.86 percent and Squarespace at 70.39 percent, both ahead of WordPress at approximately 43 to 46 percent.
Durable recorded a desktop PageSpeed score of 90 with an LCP of 1.3 seconds in independent testing. For context, the 2025 Web Almanac using July 2025 CrUX data found 48 percent of mobile origins passing all three metrics across the web overall.
One aggregate figure circulating in SEO coverage claims only 29 percent of AI-generated sites pass Core Web Vitals compared to 42 percent overall and 61 percent for hand-coded modern framework sites.
The methodology behind that figure is not fully transparent, and it should be treated as directional rather than definitive. What is clear from the platform-level data is that the major server-rendered AI builders perform comparably to or better than WordPress across these metrics, while pure SPA outputs from vibe coding tools have not been independently measured with current data.
Schema and Structured Data: What the Builders Actually Provide
Structured data helps search engines understand what a page is about and can qualify pages for rich results in search.
Wix automatically applies schema to products, blog posts, and booking pages, and added AI-generated structured data for blog content. Squarespace has auto-injected basic schema since 2016, though it covers only a limited set of types and cannot be fully customized.
Durable generates basic local business schema. Framer and Lovable do not generate schema automatically and require manual JSON-LD implementation.
One nuance worth knowing: an Ahrefs study published May 11, 2026 tracked 1,885 pages that added schema markup against 4,000 control pages in a controlled analysis.
The findings showed Google AI Mode citations increased 2.4 percent, ChatGPT citations increased 2.2 percent, and Google AI Overviews citations decreased 4.6 percent, none of which were statistically distinguishable from zero.
Google’s own May 2026 AI optimization guide explicitly lists “structured data obsession” among tactics that do not improve AI citation rates. Schema matters for rich results in traditional search. For AI search visibility specifically, it is not the lever many assume it to be.
Does adding schema markup improve AI search visibility for a website?
According to an Ahrefs controlled study published May 11, 2026, tracking 1,885 pages that added schema against 4,000 control pages, schema markup produced no statistically significant improvement in citations across Google AI Mode, ChatGPT, or Google AI Overviews.
Google’s May 2026 AI optimization guidance explicitly identifies structured data as a poor lever for AI citation rates. Schema remains valuable for qualifying pages for rich results in traditional search, but it does not meaningfully increase the likelihood of being cited or referenced by AI search tools.
The Content Problem That Affects Every Platform
Google’s March 2026 core update sharpened the Experience component of E-E-A-T more than any prior update. SE Ranking called it the most volatile core update in Google’s recorded history, with approximately 79.5 percent movement in Top-3 results.
What lost ground was thin, template-generated content with no first-hand experience attached to it. What held ground was content anchored to specific observations, named expertise, and original perspective.
This matters for AI-built websites because the default copy that ships with tools like Durable and Wix ADI is generic by design. The platform fills in your industry, your location, and your service type, and produces boilerplate that could describe any business in your category.
That copy is exactly what the March 2026 update was built to suppress.
The platform architecture is largely irrelevant to this outcome. A site on Framer with generic AI-written copy has the same E-E-A-T problem as a site on Durable. A site on Wix with copy built on deep market research, real customer language, and named first-hand experience does not.
This is the part of the AI website conversation that is critical for business owners to understand, because it requires confronting something uncomfortable…
The tool that built the site is not the problem… The intelligence underneath the content is.
Now I’m not saying the regular website or AI-built website you spent thousands of dollars on is hurting your business from getting found in search and AI search. You need to do some research on when the site was built, how it was built, and where the content came from.
I met with a business owner yesterday for coffee. She owns a successful business in The Woodlands, Texas and an expert at what she does.
But she’s not an expert on marketing or the constantly shifting world of online marketing, AI, Google and AI search, etc. Very, very few business owners are. Who has the time to keep up with all the changes happening in AI while trying to run a business? Welcome to my complex world…
So business owners rely on agencies, consultants, and others to help them with their marketing, which is great, just as we rely on experts to fix our AC or put a new roof on our house.
However, when I went through her website, I began to spot the problems. To most business owners, if content reads well, looks good, they think they are good to go. However, there is a sharp rise in people creating websites with AI and using generic AI content for the website, which is something that Google or AI will never show any love to.
There is a serious gap in AI education for small business owners. And I am seeing them being taken advantage of left, right, and center. Many of the YouTube videos about making money with AI are about selling AI services to small business owners.
Yes, there are plenty of legit videos out there that are not in it for the views or to drive clicks to their affiliate link. I’m talking about the videos that say things like… “Create a website in 2 minutes, sell it for $1,000 to a small business owner…” Or, “Create all the social media content a business needs in one month in 10 minutes and charge $750 a month.”
As a business owner, I suggest running some searches on YouTube to understand why you’re suddenly getting a lot more calls, emails, even postcards about AI agents, AI voice agents, AI websites, etc. Search for things like: “Make money with AI local business, or “AI agency business opportunity”.
Again, as someone who has been in marketing for 27 years, I want to state again that there is nothing wrong with AI agencies, those who use AI to build websites, those to sell AI voice agents to businesses, AI automations, etc.
What is wrong, in many cases, is how the opportunity is being sold and the misinformation between those who are creating and selling and what business owners are being told and charged. It comes down to education about AI and knowing the right questions to ask to these providers to ensure they are offering something they deeply understand.
Yes, an AI website builder can produce a structurally sound, technically indexed site in minutes. But what it cannot produce is a site that knows your market, knows your competitors, knows why customers choose you, and reflects all of that in copy that AI search tools and human readers can distinguish from every other site in your category.
That gap is what the hidden layer underneath AI content determines, and it is a problem that exists completely independently of which website builder you use.
If the copy on your site sounds like it could have been written about any business in your space, it was. And in a post-March-2026 search environment, Google is increasingly able to tell.
Does AI-generated website copy hurt search rankings in 2026?
Google’s guidelines state that AI-generated content does not violate policy unless it was created with the primary purpose of manipulating rankings.
What the March 2026 core update demonstrated is that thin, generic, template-generated copy, regardless of whether AI wrote it, lost significant ground when it lacked first-hand experience signals.
Content anchored to real operational knowledge, specific observations, and named expertise was not penalized. The risk for AI-built websites is not the tool that assembled the structure.
The risk is the default copy those tools generate, which is typically indistinguishable from every competitor in the same category and provides no E-E-A-T signal that would cause Google or an AI search tool to treat the page as an authoritative source.
How to Evaluate Any AI-Built Website Before You Pay for It
The View Page Source test establishes whether the site is discoverable at all. Run it on the builder’s own demo sites before committing. If the source shows content, you have a server-rendered or pre-rendered output. If it shows a script shell, you are looking at a client-side rendered application that will be invisible to AI search and may take significantly longer to index in Google.
After the rendering question, check the copy. Open the site and read it. If every sentence could describe any business in your category without changing a word, the copy is not doing any work for you. It will not be cited by AI search tools, it will not carry E-E-A-T signals, and it will not give a prospective customer a reason to choose you over the next result.
A technically sound site with generic copy has the structural foundation and nothing else.
The AI voice agent post I mentioned earlier, covered a parallel version of this same problem, where a tool that handles one operational function gets sold as a complete AI solution. What the AI voice agent sales pitch leaves out is the same thing the AI website pitch leaves out: answering the phone and getting found online are real problems worth solving, and neither one addresses what happens when someone actually evaluates your business.
That second layer is where the competitive separation happens, and it requires something neither a voice agent nor a website builder provides.
What should a small business owner check before buying an AI-built website?
Run the View Page Source test on the builder’s demo sites: right-click any page, select View Page Source, and check whether your headline, body copy, and meta tags appear in the raw source code.
If they do, the site is server-rendered or pre-rendered and indexable by Google and AI crawlers. If you see only a script shell with no content, the site is client-side rendered and will be invisible to ChatGPT, Claude, and Perplexity regardless of Google performance.
After the rendering check, read the default copy the platform generates. If it could describe any business in your industry without modification, it carries no E-E-A-T signal and will not be treated as an authoritative source by search engines or AI tools.
Where This Fits in the Broader AI Picture
AI website builders are one of seven distinct AI categories, each solving different problems and each currently being resold by people with varying degrees of actual expertise. The website category sits inside Writing and Content, with some overlap into Coding and App Building depending on the platform.
A vendor building AI websites is operating in one slice of one category. Whether their output is technically sound for search depends on their platform choice. Whether it is strategically effective for your business depends on something the platform cannot provide.
The businesses building durable competitive positions with AI right now are not doing it through AI website builders or AI voice agents alone.
They are building the intelligence layer that makes every AI tool they use more effective, because every AI output reflects exactly what the system has been given to work with.
A website built on generic training data produces generic content. A system built on your market intelligence, your competitive positioning, and your verified customer language produces something a competitor cannot replicate by signing up for the same tool.
That is what AI architecture as a competitive moat actually means in practice, and it is the variable that separates businesses using AI from businesses being represented by AI.
If your current website was built by an AI tool and you are not sure which category it falls into, the View Page Source test takes thirty seconds. If the results concern you, or if the copy problem resonates more than the technical one, ReadyStacks is worth a conversation.
Although I do not provide website services, when I work with a new client, I do a very deep dive into their existing website and online presence (using one of my own proprietary stacks I use internally). I do this to ensure they are ready to not just operate in this new AI world, but also ready to dominate their market, start generating more leads, calls, and customers from their competitors once we start using the ReadyStack I create for them.
I did one of those deep dives for an attorney last week. I ended up with seven pages of issues that was hurting them. Their AI-generated generic content was just one bullet point of many issues on their website. And from what I understand, that was after hiring another agency to fix what the previous one broke. There’s still a lot still broken and it is costing them phone calls every single day.
AI websites and the thousands of AI tools launching daily are genuinely remarkable and give business owners more opportunity than at any point in the history of marketing.
Just be sure you understand what you are buying, the questions to ask, and that whoever is building for you actually understands the current landscape, not the one from six months ago. And the king needs his own voice. Yours.
