Almost every website conversation I have at the moment includes the same question: “Should we add AI to it?” Usually that means a chatbot, an AI search box, or some kind of automatic recommendations. Sometimes it’s a great idea. Often it isn’t. This week, Smashing Magazine published a short, practical test that helps answer the question properly, and I think every business owner and designer should have it in their back pocket.

What happened

On 12 November, UX researcher Victor Yocco published a four-question audit for product teams considering an AI feature. His central idea is simple. An AI feature is only worth having if what it gives the user is more than what it costs them in effort to adjust to it and check its work. He calls this the break-even line.

That second part is the bit people forget. AI rarely removes work completely. Often it swaps typing for checking. If someone has to read every AI answer carefully to make sure it’s right, you may not have saved them anything.

The four questions

Here’s my plain-English version of his audit:

  • Does it reduce the total effort for this particular user? Count everything, including the time spent checking the AI’s output. Fewer clicks on its own doesn’t prove anything. And remember that what helps a beginner might just slow down an expert.
  • Is it better than how people do it now? If your customers can already find your opening hours in two seconds, a chatbot that answers the same question isn’t an improvement.
  • Can people spot when it’s wrong? Some mistakes are obvious. Others are invisible until it’s too late, like a suggestion to visit somewhere on the day it’s closed. If errors are hard to spot and costly, the feature needs to show where its answers come from, or it shouldn’t go live.
  • Will people still be able to tell when it’s wrong a year from now? This is the clever one. If people rely on AI for long enough, they can lose the skills they need to catch its mistakes. Yocco cites a radiology study where even experienced readers did much worse when the AI’s suggestion was wrong.

He suggests running all four questions together, before any design work starts, ideally with someone who actually does the task for a living. The answers then become part of the design brief, rather than a simple yes or no.

Why it matters for small businesses

I work on financial data and analytics products, where a wrong number isn’t just annoying, it can lead to a bad decision. That has made me fairly cautious about AI features that sound helpful but quietly add risk.

For a small business website, the stakes are different but real. A chatbot that confidently gives the wrong price, invents a returns policy or promises a delivery date you can’t meet will cost you more in trust than it saves in phone calls. On the other hand, an AI tool that helps customers find the right product from a large range might be a genuine win.

The point isn’t that AI is good or bad. It’s that you should be able to explain, in one sentence, how it makes a specific task easier for a specific person.

Try it on your own website

If you’re thinking about adding AI to your site, sit down with a cup of tea and answer these:

  • Who exactly is this for, and what are they trying to do? “Customers” isn’t specific enough. “First-time buyers comparing our three service packages” is.
  • How do they do it now, and what’s wrong with that? Sometimes the honest answer is that a clearer page would fix the problem faster and more cheaply than AI.
  • What happens if it gets something wrong? Who notices, and what does it cost you?
  • Can people easily skip it? Always give visitors a plain route to a human or a normal page.

For designers, I’d suggest adding these questions to your discovery workshops. They give you a polite, evidence-based way to push back when “add AI” arrives in a brief without a clear reason behind it.

Used well, AI can make a website more helpful. Used just because everyone else is, it adds cost, risk and confusion. A ten-minute audit before you start is a very good investment.