All my clients wanted a carousel.
Now it's an AI chatbot.

A decade of freelance web development, condensed into a single observation: the thing clients think will save their website keeps changing. I've lived through every era. I have receipts.

It starts with a Slack message. Or a Google Meet where someone shares their screen and pulls up a competitor's website. Or — my personal favourite — a PDF mood board assembled with all the energy of a Pinterest board at 11pm on a Tuesday.

And somewhere in there, every single time, is the thing. The feature. The silver bullet. The one addition that will transform their modest website into a conversion machine.

I've been doing this for over a decade. I've seen the thing change more times than I can count. And right now, in 2026, the thing is an AI chatbot.

Chapter 01A Brief, Traumatic History

Let me take you back. Back to the carousel era. Those were dark times — not because carousels are inherently evil (they are, a little), but because of the certainty with which clients demanded them. No brief was complete without a rotating banner. Hero image? Not enough. It needed to slide. Automatically. Every three seconds. Whether anyone was reading it or not.

I built so many carousels. Beautiful carousels. Responsive carousels. Carousels with parallax. Carousels with autoplay that pause on hover, as if respecting the user's attention for half a second would redeem the whole enterprise. The analytics on those carousels were devastating. Slide two: 3% engagement. Slide five: statistically zero.

Then came the era of the chatbot widget. Not AI — this was pre-GPT. These were rule-based bots that lived in the bottom-right corner of every website like an overeager intern. They had five responses. Three of them were "I'll connect you with our team!" The team was one person who checked their emails twice a day.

Then video backgrounds. Then parallax scrolling. Then "we need a quiz" (the quiz era was genuinely fun, I won't complain). Then infinite scroll. Then micro-animations on everything. Then dark mode. Then — and this one still haunts me — the brief that just said: "Can we make it more Web3?"

Chapter 02Why the AI Ask is Different (and Also the Same)

Here's the thing I have to be honest about: the AI chatbot era is not entirely without merit. Unlike carousels — which were always solving a client problem ("we have too much content and can't decide what to prioritise") rather than a user problem — AI assistants can, when implemented thoughtfully, genuinely help people find what they're looking for.

But the way clients are asking for them? That part is very, very familiar.

"We want something like ChatGPT but for our brand." Okay. What would users actually ask it? "Everything. Anything. Whatever they want." Right. What's your content strategy for feeding it accurate information? "Can't it just learn from the website?" Sort of — let's talk about RAG pipelines and hallucination risks. "Can we have it done by Friday?"

The carousel AI chatbot has become the new symbol of being digital-first. Of taking the web seriously. It signals to stakeholders that the company is Modern and Innovative and On It. The fact that a well-organised FAQ page and a decent search bar might serve users better is, frankly, beside the point.

Chapter 03What I've Learned to Do

I used to push back harder. I'd arrive at kickoff calls armed with statistics: carousel click-through rates, bounce rate data, case studies. I'd make the case for simplicity, for content strategy, for fixing the navigation before bolting on anything new. I was right, and I was also, frequently, annoying.

Now I've learned to ask better questions. Not "why do you want a chatbot?" but "what does success look like in six months?" Not "users won't engage with this" but "which user need are we solving, and how will we measure it?" The brief doesn't change, but the conversation shifts. Sometimes the chatbot still happens. Sometimes we end up rebuilding the information architecture instead, which is where the real problem usually lives.

And sometimes — I'll be transparent — I build the chatbot. Because the client is right that their customer support load is overwhelming, and a well-trained AI assistant with good guardrails and a clear escalation path actually does solve the problem. The technology, when it fits, can be remarkable. The mistake is assuming it always fits.

Chapter 04What Comes Next

I don't know what the next carousel is. A few years ago I'd have said AI chatbots — turns out I was just slow. Right now the contenders, based on the vibes coming out of certain client Slack channels, seem to be: ambient interfaces, spatial computing tie-ins, and something vague involving "agents that just handle it."

The pattern is always the same. A technology becomes culturally visible. A client sees it on a competitor's site, or in a conference keynote, or in a newsletter their CMO forwarded with the subject line "FYI 👀". A brief arrives. The feature is requested with urgency. And somewhere, a developer opens their laptop and starts googling how to actually build the thing.

I've made my peace with it. The carousel was how I learned CSS transitions. The chatbot era (round one) taught me event-driven architecture. The video background phase introduced me to performance budgets. And whatever this AI moment becomes, I'm learning things I'll carry forward.

That's the job. You absorb the trend, you understand it well enough to know when it helps and when it doesn't, and you try your best to steer the ship toward something that will still make sense in two years. Sometimes you succeed. Sometimes you ship a carousel.

Either way, the work was real, and the lessons were too.