Sometimes you shouldn’t. AI and the newer site builders can get a decent-looking website live in an afternoon for very little money, so the build itself is no longer the expensive part. Decide by two things: how much the site affects whether buyers trust you, shortlist you or contact you, and whether you have someone who knows how to make it do that job. If the site mostly needs to give people accurate information, an AI builder may be all you need. If it shapes buying decisions, someone has to make informed calls about the message, the proof and the experience. That can be your own team, an outside advisor or an agency. Pay an agency for the whole project when you also need one team to own the build, the migration and the launch.
This is for owners, CEOs and marketing leaders deciding whether a website project is still worth funding now that the tools have changed. I run a marketing agency that builds websites and I advise companies on website decisions, so I have a stake on both sides of this question.
What AI changed about websites, and what it didn’t
The cost of getting to a first version collapsed. A person with no design or development background can describe what they want and have a working site the same day. I build with AI every day, and the speed is real.
The look of a site also tells a buyer less than it used to. A polished, professional site used to take real time and money, so it was a reasonable sign that a serious company was behind it. Now that it’s cheap to produce, good looks alone give a buyer less reason to assume that. In the sites I review, I see more and more that look polished and say almost nothing specific about the business.
What didn’t change is where the answers come from. AI can help analyze customer interviews, compare competitors’ sites and draft possible messages, and it’s good at it. It needs true information to work from, and someone still has to judge whether its conclusions hold up. If nobody has talked to customers, looked at lost deals or worked out what actually makes the company different, AI will produce confident, convincing answers without any way of knowing whether they’re right.
When building it yourself is the right call
The real test is how much the site has to do in the sale, and what it costs you if it does that badly. Company size and budget matter less.
Build it yourself, or with a builder’s AI, when:
- The site just has to exist. A holding page, a landing page for an announcement, a site for a venture you’re still testing. A client of ours needed a one-page site live in time for an announcement and built it with AI. It needed a few decisions from us, mainly about where the announcement should point, not a build.
- The money is better spent somewhere else right now. An owner we work with was opening a new location while still deep in startup costs. He told us he’d be fine with an AI-generated site from his hosting company instead of a custom build. For a new store with that much money already out the door, that’s a reasonable call.
- Your current site mostly works. Sometimes the right spend isn’t a new site at all. Fixing a few weak pages on a site that’s doing its job can beat any rebuild, AI or otherwise.
Be careful with one common assumption: that a referral business doesn’t need its site to do much. It may only need to confirm that you’re credible and capable. But a referred prospect may still check your site against a competitor’s before calling. Before you decide, ask your sales team what prospects look at before they get in touch, and ask a few recent customers whether the site played any part. You may find it matters more, or less, than you assumed.
What you’re paying for when the site has to win a buyer
Once buyers use your site to decide (comparing you against the two other tabs they have open, checking whether you’ve done this before, deciding whether to fill in the form), the question changes. The build is still cheap; what costs something now is deciding what the site should say and prove.
These are the questions I ask of a site that has to carry a buying decision:
- “Why would a buyer pick us versus the two tabs sitting open next to him?”
- “Which of all the claims on the website will actually read as proof?”
- “Does the human experience of actually using the website result in something that’s going to move a business objective forward or not?”
- “Does all the design work, and the design decisions that have been made in the website, actually support something real rather than just decorative work?”
Answering them well produces things you can see: a clear order of priority for who the site is for, a page-by-page plan of what each page has to do, a list of the proof you need to gather (case results, credentials, the specifics a buyer will check), and the tests the finished site has to pass before launch. If someone is charging you for website strategy and you can’t point to those, ask what you’re getting.
Those answers live in your sales conversations, your win and loss history, your margins and the objections your team hears every week. Whether a person or a tool lays out the pages matters much less than whether someone has done that digging.
The part nobody prices: keeping it working
The other cost shows up after launch, and it’s easy to miss when you’re comparing a near-free build with a five-figure proposal.
I’ve tested this myself, rebuilding sites with AI coding tools. They came together fast, and one scored a perfect 100 on Google’s speed test. But they were hard to work with for anyone who doesn’t write code, because there was no simple way to edit a page. That doesn’t fit how my team and I actually run websites, so we stay with sites we can update without a developer.
Before you commit to any build, AI or otherwise, ask:
- Who will edit it? If the answer is “whoever built it” and that person is a tool or a contractor you can’t reach, every small change becomes a project.
- Who owns it? The domain, the hosting, the analytics and the logins should be in your company’s name, whoever builds the site. That’s as true of a cheap build as an expensive one.
- Does it actually work? Forms that reach the right inbox, tracking that records what you think it records, pages that work on a phone. Make sure someone has tested all of it, however the site was built.
- What happens to what already works? If you’re replacing an existing site, the pages that bring in search traffic need to keep their addresses or be redirected properly. A fresh build that ignores the old site can quietly throw away years of search visibility.
- What happens when the tool changes? A site that only one platform can edit is only as stable as that platform’s pricing and roadmap.
These questions decide whether a cheap build stays cheap. Count your own people’s time too: an AI builder’s subscription is small, but the writing, editing, testing and upkeep still land on someone.
Your options, side by side
| Situation | Likely best fit |
|---|---|
| A simple information site, little harm if it’s imperfect | Build it yourself or with an AI builder |
| The site matters, and you have capable marketing and technical people in-house | Your own team, using AI |
| The site matters, you can build it, but the message and structure are unclear | Your team builds; an outside advisor shapes the message, the brief and the final review |
| The site matters, and you’re missing the expertise or the capacity across several pieces (strategy, design, build, migration, testing) | An agency that owns the whole project |
| The current site works reasonably well, but the message or conversion is weak | Improve what you have before commissioning a rebuild |
These are starting points, not rules.
The middle row is where I’d look first if you have someone who can build and maintain the site. I sell that kind of advisory work, so factor that into my recommendation. If you don’t have that person, splitting the advice from the execution can make the project harder and more expensive, and I’d rather see one team own the outcome. An agency-led project also makes sense when a failed launch would be costly and there’s no one inside to coordinate the moving parts.
Common questions
Can AI build a website for a B2B company?
Yes. It can build a competent site quickly, and for a site that mainly has to exist, that’s often enough. What it can’t do on its own is know your buyers, your proof and what sets you apart. When the site is part of how deals get decided, someone has to supply that and check the result.
Is an AI-built website worse for SEO?
On its own, no. Trouble comes from generic content, technical problems that stop pages being indexed, weak internal links and a badly handled move from an old site. The same mistakes happen on agency-built sites; a fast build just makes them easier to skip past.
What does a website cost if we don’t build it ourselves?
It depends far more on what’s included than on who builds it. I’ve written about why two quotes for the same site can be ten times apart in how much a website should cost.
If you’re deciding whether your site needs a rebuild, a restructure or a better brief for whoever builds it, that’s what my website consulting work is for. If you’d like to talk it through, start on Work With Me.
