Open a handful of small business websites recently made and look at what sits above the main heading: little “eyebrow” headings above the H1, the same layout on every site, and any “How it works” section numbered with “01, 02, 03”, etc.
These sites were built with AI, and anyone who has spent time online over the past couple of years can spot them in seconds. The sections fade in on the same scroll animation, and the copy promises to elevate and streamline your business in the same cadence as the last site you visited.
The appeal is easy to understand. You can describe your business to an AI tool in a paragraph and watch a website appear seconds later—it’s like magic! And, yes… I wrote that em dash myself...
If an agency has quoted you a few thousand dollars, the obvious question is why you'd pay for something a chatbot will hand you for free. AI builds what you tell it to build, so the result can only ever be as good as the direction it's given. Garbage in, garbage out. A website that brings in leads needs someone at the controls who understands how it should be structured and why.
AI defaults to the average
Language models learn from enormous volumes of existing material, including millions of websites. When you ask one for a website without specific direction, it produces the most statistically likely website, which is the average of everything it has seen. Every gap you leave in the brief gets filled with the most common answer, so the less you specify, the more your site looks like everyone else's.
Average is a problem for a business website because the site's job is to win customers who are comparing you with your competitors. If your site is interchangeable with theirs, you've given a visitor no reason to choose you.
Big prompts produce fragile websites
Every AI model has a context window, which is the amount of information it can hold and work with at once. Ask it to design and build a full, responsive, multi-page website in one go and that window fills quickly with code and the model's own earlier output. Chroma's research on context rot (opens in a new tab) tested 18 leading models and found their reliability drops as the amount of input grows, even on simple tasks. Past a certain point the model loses track of decisions it made earlier and starts inventing details to fill the gaps.
The consequences tend to show up after launch. A layout that holds together on a laptop collapses on a phone. A contact form looks finished but never delivers an enquiry to your inbox, buttons link nowhere, and somewhere in the code there may be a security vulnerability that nobody reviewed because nobody knew to look.
You usually find these faults the way your customers do, through enquiries that never arrive.
A working website still needs a home
Suppose an AI tool did produce a site that looked distinctive and worked on every device. You'd still need to work out where to host it, connect your domain and configure the DNS correctly, put protection in place against DDoS attacks and spam, and set up monitoring so you know the moment the site goes down.
Launch is where the ongoing work begins. Dependencies need updating as browsers and frameworks move on, and your content needs to change as your business does. When something breaks on a Friday afternoon, someone has to know where to look and how to fix it without breaking something else.
Use AI the way you use Google
AI is a groundbreaking piece of technology and a powerful productivity tool when it's used well. I think the closest comparison is what Google did to the encyclopaedia. Google took the slow, manual job of hunting through reference books for an answer and turned it into a search that takes a fraction of a second.
You still have to know what you're looking for, and you need a basic understanding of how search works to get anything useful out of it. Type "what's my Facebook password" and you'll get nothing helpful, because Google can't retrieve something it was never built to know. Type "how to reset my Facebook password" and you'll have the answer in seconds. The tool is the same in both searches, and the person using it makes the difference.
AI works the same way. To get a website that brings in business, you need to understand how AI works and, more importantly, how the thing you're asking it to build works, so you can guide it through the exact structure and architecture your site needs. This is why professional web developers are still in high demand, because it takes a skilled software developer to know the ins-and-outs of the code that’s being written—whether by hand or with AI—and how the underlying structure of the website should function.
How we build websites at Liquid Lava
Every project starts with a conversation about your business. We talk through who your customers are and what a good enquiry looks like for you, then design your site around established principles of layout, hierarchy, customer journeys and search visibility. Each element on the page earns its place by helping your business generate leads or sales.
Once you've approved the design, we build it to be responsive across all devices and functional for your business. AI tools help us move through the repetitive parts of that work faster, the same way a search engine speeds up research, while every decision about the structure and code of your site stays with our team.
Once your site is live, we manage hosting and look after updates, uptime monitoring and fixes, so you never have to think about DNS records or dependency updates. Your copy and photos sit in an easy-to-use content editor, so you can make changes yourself whenever you like.
If you're weighing up whether to build your site yourself with AI, book a free 30-minute discovery call with us. We'll talk through what your business needs from its website and how we'd build it.
AI has made it quicker than ever to put a website online, and easier for your customers to tell when nobody thought it through. A site designed around your business by people who understand it is what earns their trust.
What goes in still decides what comes out.




