- Home
- About
- Service
- News & Insight
- Contact
- Resources
Innovative Thinking
-
Innovative Case
-
You are not paying for the pages. An AI website builder will produce a full set of them in the time it takes to describe your business, and what comes back is usually competent. The AI website builder vs web agency question is really about everything either side of that moment: what the site is for, which pages already earn their place, what to keep from the site you have, and who answers when the enquiry form stops delivering.
Production has become cheap. Judgement, evidence and accountability have not. An agency that only assembles pages is in genuine trouble; one that decides what the pages should be is not. Here is what that means for a business choosing between regenerating its website and revamping it.
Start with what is true: the case for an agency does not rest on the tools being bad. They are not. Describe a business in three sentences and a current builder returns a coherent homepage, service pages, a contact page and the legal boilerplate, in one consistent layout, with copy that reads like English. A few years ago that was a fortnight of work.
They are strongest at the blank page and at repetition: first drafts, alternative layouts to react to, a privacy page nobody needs to be original about. If you have no website at all, a generated site beats waiting for the budget to do it properly.
What a builder cannot do is know anything about your business that you did not tell it. Most of what makes a website earn its keep is exactly that.
The builder is rarely the weak part. The brief is.


A builder answers one question very well: what should this page look like? It cannot reach the questions that come before it. Which of your forty pages produced last month’s enquiries? What did the last three people who did not enquire get stuck on?
Those are not design questions, and no prompt reaches them, because the answers live in your analytics, your inbox, your WhatsApp history and the heads of the people who answer the phone. Gathering them is slow, unglamorous work, and it is the part most likely to be skipped — by an agency as easily as by a tool.
So the fair test of any quotation, ours included, is whether it describes decisions or only production. A proposal listing pages, sections and revision rounds is a production quote, and production is what has become cheap. One that opens with what will be measured, kept and removed is selling you something a builder does not.

These are the points where a website is won or lost. None is a design task.
The most common worry is the least justified. Google does not penalise a page for having been generated. Its guidance on AI-generated content states that using automation, including AI, to generate content with the primary purpose of manipulating ranking in search results is a violation of its spam policies — the condition is the purpose, not the tool. Its current documentation, updated in December 2025, adds the instruction that matters: focus on accuracy, quality and relevance, especially when automatically generating content.
The real risk has a name. Google’s spam policies define scaled content abuse as many pages generated for the primary purpose of manipulating search rankings and not helping users, and name generative tools used to produce many pages without adding value as one route there. In practice that is the thirty near-identical service-area pages spun up in an afternoon, not the five pages that answer what your customers ask. Neither the accuracy check nor the restraint is a setting inside a builder. Both are somebody’s job.
Generated sites are rarely wrong on the day they launch. They are unfinished in ways nobody looks at until something depends on them.
Accessibility is the clearest case — less an AI problem than a nobody-checked problem that speed makes larger. WebAIM tested the top one million home pages in February 2026 and found detectable WCAG failures on 95.9% of them, averaging 56.1 errors per page. The commonest faults were low-contrast text on 83.9% of pages and missing alternative text on 53.1% — defaults nobody overrode. Anything that produces pages quickly produces those defaults quickly too.
The second cost is verification. Google’s DORA research into AI-assisted software development — engineering teams, not websites, but the mechanism carries — reported in 2025 that higher AI adoption raised throughput and instability together, and that time saved during generation is often re-allocated to verification overhead. It describes AI as an amplifier of what an organisation already does, well or badly. A business with nobody reviewing the output collects the volume rather than the saving.
The third is ordinary upkeep, which decides what a site costs over five years rather than five weeks: speed on a mid-range phone over 4G, plugin and platform versions, forms that quietly stop delivering, content that ages. We have written separately on what a maintenance plan should cover and on the signs that upkeep has become the main expense. A generated site arrives with none of it arranged.
Most businesses reading this are already in that situation, and it changes the answer. An older site is not a blank page: the history, the ranking pages and the real photographs are already there, along with a Google Business Profile pointing at specific URLs. Regenerating discards all of it, and then you pay to earn it back.
The sensible order is unchanged by any of this. Find out what the current site earns, page by page. Keep what earns, and the URLs it earns on. Rebuild the structure, the mobile enquiry path and the service pages around what you kept. Then use whatever tools are fastest for production inside that plan — we do — because production was never the expensive part.
Yes. Google judges the page rather than the tool that made it, and says so plainly. What stops generated sites ranking is more ordinary: service pages that answer nothing specific, no internal linking worth the name, and a structure taken from a template rather than from what customers search for.
Not necessarily, and starting again would be. Treat it as a first draft that exists: check every factual claim against the business, fix the enquiry path on a phone, rewrite the two or three pages that should be bringing work, and add the specifics a generated page could not invent. That is a revamp, and it costs less than a rebuild.
No rule requires it. Google suggests adding information about how content was created where that makes sense for your audience — a judgement about trust rather than a ranking tactic. What matters more is that anything presented as your own experience genuinely is yours.
Not the pages. We use the fast tools where they save time, rather than pretending a homepage takes three weeks to assemble. The fee covers the work either side: finding out what the current site earns before anything is redesigned, deciding what to keep, writing the parts only your business knows, and answering when something breaks. You can read how we work, and our web development and design service begins with that audit rather than a layout. For most of the sites we are shown, the recommendation is a revamp — not a rebuild, and certainly not a regeneration.