A small business website now serves two audiences at once: the people who visit it, and the AI assistants that customers ask before they visit. Most sites work for one but not the other. I build sites where the text is real text, the data stays consistent, and there is nothing in it just because a platform expects it there. That is how a customer's AI assistant describes you correctly instead of leaving you out.
A customer asks ChatGPT whether you exist and what you do before they visit your website. That assistant reads your site the same way a search engine does, but it reads it differently than a person does. When I talk to small business owners about commissioning a website, most of them ask about platforms and costs. What they should ask is whether the site will describe itself correctly to an AI assistant. That difference changes everything about how I build.
Why your platform choice matters less than you think
I build fast custom websites for service businesses. Most of them run on lightweight PHP, deployed automatically from GitHub, no heavy content management system behind them. The reason is not that no-code builders are bad. I choose this path because when I start from what the business needs to do, I end up building something spare and purposeful. There is nothing in it because a platform ships with it. There is nothing in it that does not serve the business or the people who find it.
But this is not really a post about platforms. It is about a shift in what works. Your website used to serve one job: let people who search for you find you. Now it serves two. It has to describe itself correctly when a person reads it and when an AI assistant reads it. Those are not the same thing, and most sites fail at one or the other.
A no-code builder can do this. A custom site can do this. What matters is not the tool. What matters is what I do with it.
What AI assistants read
When you visit a website, you see the layout, the color, the images. When an AI assistant reads your website, it sees something different. It sees the text. It sees the structure. It sees the metadata that describes what the text means. Everything that matters has to be real text in the page, not buried in an image or hidden in an admin panel somewhere. If you only put information inside a graphic or a PDF, an AI assistant does not see it.
I learned this while building our own site. Everything that matters is written as text: what we do, how an engagement works, when you can reach us, how to contact us. Nothing important hides inside a banner graphic. The trade is that I have to design with type and layout instead of leaning on a big visual. The gain is that a crawler reads exactly what a visitor reads. What you see is what an AI assistant describes.
The hidden problem that kills accuracy
Here is something smaller that most people miss, but it breaks whether an AI assistant describes you correctly. On our website the FAQ and the structured data that machines read come from a single list in the code. Both print from the same source. When someone fixes an answer on the page, the machine-readable version updates at the same time. They cannot drift apart.
If you edit your FAQ on the page and your markup lives somewhere else, they diverge. Someone updates the answer. The structured data keeps saying the old thing. An AI assistant quotes an answer that is no longer true. You look wrong.
This is one small decision, but it is the kind of thing that decides whether an assistant describes you correctly or leaves you out. It is not about the platform. It is about refusing to create multiple sources of truth for the same information.
What to ask when you hire someone to build your site
When I talk to people about building or redesigning their site, most vendor conversations go the same way. How much does it cost. What platform do they use. How long does it take.
I ask different questions instead. I ask how the text will be structured so a crawler can read it. I ask where the business information lives in the code and whether it stays consistent. I ask what happens when you update a detail: does it update everywhere at once, or do you have to find it in three different places and fix it each time.
Ask whether anything important only lives inside an image or a PDF. Ask how fast the site loads and what they do to keep it that way. Ask what breaks, and how often they have to patch things. Heavy platforms break more often. Light sites break less.
These questions sound technical. They are not. They are business questions. A site that an AI assistant can read correctly is a site that a customer finds. A site that is slow or breaks often costs you time you do not have. A site where the same information lives in three places is a site where mistakes happen.
The decision that matters
When a customer's AI assistant describes your business, you want that description to be right. You want the hours to be correct. You want the services listed to be the services you actually offer. You want the assistant to say you exist.
That happens when I build with intention. Not intention in the design or the color. Intention in the structure. Intention in refusing to hide what matters. Intention in keeping one source of truth. The platform is a tool. What I do with it is the choice.
