Your website was good three years ago. Nothing about it has changed since, and that is precisely the problem.
Enquiries are slightly down. The same three questions keep arriving by email. Someone mentioned it felt a bit dated and you were not sure whether they had a point or were just being fashionable. Nobody has complained loudly enough to force a decision, so no decision has been made.
That is what a decaying user experience looks like from the inside. It almost never announces itself. It shows up as a slow leak in the numbers, and by the time it is obvious it has usually been costing money for a year or more.
Here is why it happens, how to tell whether it is happening to you, and what to do about it that is not an expensive rebuild.
Published April 2022. Rewritten August 2026 to cover what AI has done to the pace of change, and to add the diagnostic and the fix.
Why it gets worse when nothing changes.
There is a principle in user experience called Jakob’s Law, named after Jakob Nielsen. It states that people spend most of their time on other websites, and therefore expect yours to work the way those other sites work.
Read that again, because the commercial implication is easy to miss.
Your site is not judged on its own terms. It is judged against every other site your customer used that week. Their bank, their supermarket, their airline, whatever they were looking at ten minutes before they arrived on yours. Those companies are all running their own research, spending their own money, and quietly moving what people consider normal.
So your site gets worse without anyone touching it. Nothing about your product declined. The comparison moved.
This is the part most business owners find genuinely annoying, and I understand why. You paid for something, it worked, and now it does not, and nobody did anything wrong. But it is the single most useful thing to understand about digital spend, because it explains why a website is not a purchase. It is closer to a vehicle. It works well for years, and it needs servicing to do it.
What decays and what does not.
Not everything you paid for is losing value at the same rate, and knowing which is which is where the money is.
Some things never expire. Bigger means more important. Dark text on a light background is easier to read. A button has to be large enough for a thumb. These are not trends, they are human wiring, and they have not moved in forty thousand years. A neanderthal would have tapped a button with the same accuracy you do, because the hand has not changed. What has changed is that the neanderthal would have had no idea what an app was.
So there are two halves to what you own.
The durable half is your brand, your photography, your written content, your typography and the underlying structure of the thing. Money spent there compounds. It is still working in five years.
The perishable half is patterns and conventions: how your navigation behaves, how your forms are laid out, how your checkout flows, what people expect to happen when they tap something. Money spent there is genuinely valuable and genuinely temporary, in the same way that money spent on stock is.
Most businesses I meet have the ratio backwards. They spend heavily on the perishable half every four or five years in one enormous lump, and almost nothing on the durable half in between. That is the most expensive available order.
AI has made this happen faster.
Everything above was already true. What has happened over the last three years is that the clock has sped up, in four ways that affect what your customers expect before they ever reach you.
People now expect to ask, not to search. A search box used to promise keyword matching. A growing number of your customers will type a full question into yours and judge it on whether it understood them. Most site searches do not, and the person leaves without telling you.
Answers arrive before the click. When someone searches for what you do, they increasingly get a summary rather than a list of links. Fewer people reach your site, and the ones who do arrive later in their thinking and expect more from the page. Traffic being down is not automatically a marketing problem.
The same action no longer always gives the same result. Anything with AI inside it behaves differently each time it runs. That quietly breaks the oldest assumption in interface design, which is that people learn a thing once. If you are adding AI features to your business, this is the part that needs designing, not the clever bit.
And everyone else is shipping faster. The cost of building a website has collapsed. Your competitors are updating more often than they used to, which means the baseline your customers measure you against is moving quicker than it ever has. A maintenance interval that felt sensible in 2022 is now too long.
This is why I talk about direction rather than automation. The tools have made building fast. They have not made judgement fast, and almost every bad experience I am asked to look at now sits in the gap between those two things.
The three stages every customer goes through.
People move through three phases with any website, product or service. Discovery, comfort, then boredom.
Discovery is the expensive phase. They do not know where anything is, they need support, and this is where most of them leave. Comfort is where you make your money. They know how it works, they need almost nothing from you, and they come back. Boredom is where they start looking at what else is available.
The whole job is to move people out of discovery quickly and then keep them in comfort for as long as possible. That does not mean never changing anything, because comfort turns into boredom on its own. It means small, frequent, easily absorbed improvements rather than occasional dramatic ones.
This is the commercial argument against the big redesign, and it is worth stating plainly. A full rebuild takes every customer who was comfortable and puts them straight back into discovery, the phase where they need support and are most likely to leave. You pay for the rebuild twice: once to the agency, and once in the churn it causes.
Six signs your UX has gone bad.
Decay shows up as a business problem before anyone identifies it as a design one. You can check all six of these against information you already have, without commissioning anything.
- The same few questions dominate your inbox. If three or four questions account for most of your inbound support, those are not support problems. They are things your website should be answering, and a person is being paid to answer them instead.
- People drop off at one specific point. Losing people steadily across a whole process usually means the offer is wrong. Losing them at one step almost always means something is broken there, and it is often an afternoon of work once you know which step.
- Customers ask for things you already do. This is the clearest signal available that your site is hiding something, and the worst possible reason to go and build something new.
- Visits are getting longer but sales are not. Time on site is only good news when people are enjoying themselves. The rest of the time it is a measure of difficulty.
- Your own team has workarounds. If the people who work there navigate by memory and shortcuts, a first-time customer has no chance at all.
- Nothing has changed in eighteen months. Given everything above, a site untouched for that long has decayed whether or not anyone has said so. Silence is not a pass.
None of these need research. They need somebody to look.
How to fix it.
The instinct is to rebuild. Resist it for a moment, because a rebuild resets the clock and starts the same decay again, and it is usually the most expensive way to solve a problem you have not diagnosed yet.
Find the break before you spend anything. Your analytics will tell you where people leave. They will not tell you why. Take the where from the numbers, then sit five people down and watch them try to do the thing. Five is enough to find most of it. It is a morning, not a research project. And resist fixing the first thing you notice, because the first thing you notice is rarely the thing costing you money.
Fix the drift, not your taste. The question is never whether you like a pattern. It is whether it still matches what your customers meet everywhere else. If your navigation, your forms or your checkout behave differently from what has become normal, that difference is now costing you, and it needs a better reason than it being how you have always done it.
Put the money on the durable half. Brand, content, photography, structure and clarity keep paying. Patterns and conventions need maintaining. A design system is worth having for exactly this reason: it separates the two for you, so the durable part sits in one place and the perishable part can be updated without touching it.
Then make maintenance normal. A scheduled review, a short prioritised list, and a few things shipped from it regularly. Small and continual beats large and occasional, because it keeps customers in the comfort phase instead of throwing them back into discovery every four years.
What this means for your budget.
Every website begins losing ground the day it launches, no matter how much you spent or who built it. That is not a criticism of your agency. It is how the thing works.
The businesses that get this right are not the ones spending the most. They are the ones spending steadily. They put the bulk of their money into the parts that do not expire, and they keep a modest, regular budget for the parts that do.
The ones that get it wrong spend nothing for four years, then spend a great deal all at once, lose a chunk of their existing customers to the disruption, and start the same cycle again.
AI has not changed that argument. It has shortened the interval. The baseline is moving faster than it ever has, which makes neglect more expensive and maintenance better value than it was three years ago.
If any of the terms here were unfamiliar, most of them are explained plainly in the UX and AI glossary. If you would rather someone else did the looking, that is what design is for.
