Web design in the AI era: how we avoid AI slop
The short answer
AI slop is what happens when a website is generated instead of designed. Same layout, same stock hero, same three column feature grid, copy that says nothing. It is fast and it is cheap, and it looks exactly like everyone else's. We avoid it by keeping the process the same as it has always been. Discovery, planning, pen and paper, Figma, a real build, launch, then growth. Over ten years of doing that is what makes the difference, not the tools.
What AI slop actually looks like
You have seen it, even if you have not named it.
The hero section with a phrase like "empowering your digital journey". Three icons in a row that could belong to any business in any sector. A stock photo of people in a meeting room in a city you have never been to. Copy that mentions "seamless solutions" and never once mentions what the company actually does.
It is not badly made. That is the problem. It is competently made and completely empty. Nothing on the page could only be true of that one business.
For a hotel in Portrush or a tour operator on the Causeway Coast, that is the worst possible outcome. The whole reason someone books with you instead of a chain is that you are not the same as everything else. A generic website argues the opposite case on your behalf.
Why this is a business risk, not a taste problem
Google formally defined scaled content abuse in its March 2024 spam policy update, and that definition still governs in 2026.
Two things matter about it.
First, the policy is method neutral. Google does not penalise content for being AI generated. It penalises content produced at volume mainly to rank rather than to help anyone, whether a person, a model or a template made it.
Second, enforcement has been getting sharper. Google's March 2026 update leaned heavily on scaled content abuse, and third party analysis of the damage pointed at repetitive AI pages, scraped rewrites and location templates with no original information in them. A further spam update rolled out between 18 and 21 August 2026.
There is also a newer angle worth knowing. On 15 May 2026 Google rewrote the opening line of its spam policies to include attempts to manipulate generative AI responses in Search. Answer engines are now covered by the same rules.
The pattern is consistent. Volume without value gets you nowhere. Specific, differentiated, genuinely useful pages are what survive, and they are the ones AI answer engines cite.
What ten years actually buys you
Judgement. That is the honest answer.
We have worked on brands including Adobe, Jack Daniel's, Jim Beam, Maker's Mark and Whiskas, and on independent businesses right across the north coast. You can see a cross section of that in our work. Very different budgets. The same underlying question every time. What is true about this business that is not true about anyone else, and how do we build the site around that.
You cannot prompt your way to that answer. It comes from sitting with an owner for an hour and listening to how they talk about their own work. The phrase they use without thinking about it is usually the headline.
Experience also shows up in what you leave out. Knowing that a page needs less, not more. Knowing which animation earns its place and which one just slows the site down. Knowing that the thing the client is most proud of is often not the thing the customer needs to see first. Those are ten year decisions, made in seconds.
The process that has not changed
Seven stages. Same as they were before any of this, and the reason the output does not look generated.
Discover. We start with the client, not a template. What the business actually does, who books or buys, what is currently getting in the way. For hospitality clients that usually means the operator's own language about their place, because that language is the brand.
Plan. We work out the right solution before we work out what it looks like. Sitemap, page purposes, what has to be editable, what the site needs to do that the current one cannot. Content comes before build, always. A beautiful shell with nothing in it is a slower kind of slop.
Sketch. Pen and paper first. Every layout starts on a page in front of us. It is not outsourced and it is not AI planned. This is the stage where the odd idea shows up, the one that would never survive a template. Rough sketches are cheap. Bad decisions in code are expensive.
Design. Into Figma, and the site is designed as a one off. Real type choices, a real colour system, real hierarchy built for that one business. See how that plays out across our branding and web design work.
Build. Real people building real things. We pick the stack to suit the job, whether that is a custom WordPress theme built from scratch or a headless Next.js build. Hand coded either way, tested on real devices, accessible, fast. A developer who understands why a decision was made can defend it when something changes. Generated code cannot.
Deliver. Review, revise, launch. We walk the client through the CMS so they own their own site rather than ringing us every time a price changes.
Grow. SEO and AEO after launch. Structured data, location pages with genuine differences between them, content that answers the questions people are actually typing into Google and into ChatGPT. This is where a lot of agencies stop caring. It is where the site starts earning.
What that looks like in practice
Two projects show the difference between a generated site and a designed one.
Portrush Surf School
Portrush Surf School is a custom WordPress build with ecommerce, and it is the clearest example we have of the process feeding directly into search.
A template would have given them one lessons page with a list on it. Instead every session type has its own page. Group lessons, VIP kids surf camp, hen and stag parties, corporate days, private lessons, family sessions, paddleboard, giant SUP, sunset SUP, bodyboard. Ten pages, each with its own booking route.
That structure was a planning decision, not a design flourish. Somebody searching "hen and stag parties in Portrush" and somebody searching "kids surf camp Northern Ireland" are two completely different people with two completely different questions. One page cannot answer both well. Ten pages can.
The same logic runs through the content. Guides on the best surfing spots in Ireland, what to do in Portrush in summer, hen and stag activities. Real answers to real questions, written by people who actually run surf lessons off that harbour.
That is why the site performs the way it does in Google and in AI answers. [FIGURE NEEDED: your actual ranking positions or traffic change, plus timeframe. For example "first page for surf lessons Northern Ireland within four months" or "organic sessions up X percent year on year".] Answer engines cite pages that are specific, well structured and clearly authored by someone with first hand experience. Ten distinct service pages and a genuine content library give them a great deal to work with. A single generated page gives them nothing.
New Row Dental
New Row Dental is a Coleraine dental practice that delivered excellent care behind a dated site with almost no visibility in local search. Patients struggled to find them and struggled to book.
We rebuilt it as a custom WordPress site with on site video and photography of the actual practice and the actual team. No stock. The content was restructured to separate cosmetic, general and emergency care, with local SEO content tuned to terms like Invisalign Coleraine and North Coast dentist. Calls to action went exactly where patients needed them.
The results have gone further than a website brief normally does. The practice has grown to the point where they have had to expand to keep up with bookings and enquiries. [FIGURE NEEDED: what the expansion actually was, for example an extra surgery room, an additional dentist, extended hours. And a booking or enquiry number if you have one. Get this confirmed by the practice before publishing.]
That did not come from a clever headline. It came from the discovery conversation, where it became clear that their most valuable service was the one the old site buried.
So do we use AI at all
Yes, and we would rather say so than pretend otherwise.
We use it the way we use any other tool. Drafting, research, checking our own logic, speeding up the boring parts of a build. What we do not do is let it make the decisions.
The distinction is simple. AI is very good at producing something plausible. It has no opinion on whether the thing is right for your business, because it has never met your business. A tool in the hands of someone who knows what good looks like is a shortcut. The same tool in the hands of someone who does not is a slop machine.
Every word, layout and line of code that leaves lowe design has been decided on by a person who can explain why it is there.
Frequently asked questions
What is AI slop in web design?
AI slop is generic, mass produced website content and design. Identical layouts, stock imagery, vague copy that could describe any business in any sector. It is usually competent and always forgettable.
Does Google penalise AI generated websites?
No. Google's spam policies are method neutral. What gets penalised is scaled content abuse, meaning pages made mainly to manipulate rankings rather than help users, whether AI, a human or a template produced them. Google defined this in its March 2024 spam policy update.
Does lowe design use AI?
Yes, as a tool inside a human process. We use it for drafting, research and speeding up repetitive work. Strategy, design decisions and final copy are made by people. Nothing is published that we cannot explain.
Why start with pen and paper?
Sketching is the cheapest place to be wrong. It produces ideas that templates and generated layouts never reach, and it keeps the design tied to that one client rather than to a pattern library.
Is a custom website worth it over a template?
It depends on the business. If you compete on being different, a template argues against you every time someone lands on your homepage. Hospitality, tourism and independent retail feel this most sharply. Portrush Surf School and New Row Dental are both examples in our work.
How long does a custom website take?
Discovery to launch typically runs six to ten weeks depending on the size of the site and how quickly content and photography come together.