AI Website Builders Won the Small-Business Vote: 53% Use Them — but Only 9% Shipped the First Draft

A new Bluehost survey of 201 US small business owners, reported October 16, puts a number on something we’ve all been watching happen in real time: 53% of them built their business website with an AI website builder instead of hiring a developer or an agency. The owners who went the AI route reported saving an average of $850 and about 32 working hours. Satisfaction ran high — 72% expressed strong approval of the result, and 97% were at least moderately satisfied.

Pair that with Freelancer.com’s numbers and the picture looks like a rout. Website design job ads on the platform fell 59% between Q2 2024 and Q3 2026. PHP gigs fell by the same 59%. WordPress assignments dropped 58%, HTML programming 55%, graphic design 43%.

Read those two data points together and the obituary writes itself: the freelance web developer is done. Except the same Bluehost survey buries two numbers that complicate the story. Only 9% of owners published the AI’s first draft without changes. And setting up the domain or hosting was one of the tasks owners were least likely to hand to the AI — only 21% did.

So 53% build with AI, 97% are satisfied, and 91% still changed the output before it went live. All three are true at once. The question is what that actually means for a business owner deciding today — and for the developers watching their job boards thin out.

What the $850 and 32 Hours Actually Buy

Let’s be honest about what most small business websites are. A plumber’s site. A bakery’s site. Five pages: home, services, about, reviews, contact. The $850 in savings is real because the thing being replaced was genuinely overpriced for what it was — a static brochure site that a freelancer knocked out in a week using a template, charged four figures for, and never touched again.

We’ve built sites both ways for clients over the years, and the AI builders are legitimately good at this specific job now. They produce a coherent layout, write passable service descriptions, and get the mobile version right more often than the average $500 freelancer did. The 32 hours saved are the hours nobody misses: picking a template, arguing about stock photos, waiting three days for a “we’re working on it” email.

The savings stop being real the moment the site needs to do something. Booking with real availability rules. A quote calculator. Checkout that doesn’t leak. An AI builder will happily generate all of these and present them with total confidence. Whether they work is a separate question — and it’s the one the 91% who edited the draft were answering.

The Work the AI Still Can’t Do

The 21% domain-and-hosting number is the most honest stat in the whole survey. Connecting a domain, setting DNS records, and pointing them at hosting is unglamorous plumbing that fails silently and expensively. Get it wrong and the site is down, the email doesn’t arrive, or — the classic — the AI builder “helpfully” registers the domain under its own account and the owner doesn’t actually own their own web address. We’ve seen that exact situation twice this year, both times discovered when the owner tried to leave the platform.

Here’s the checklist we’d run on any AI-built site before it represents a real business:

Own your domain outright. Register it yourself at a registrar you control — not inside the builder’s checkout flow if you can avoid it. If the builder relationship sours, the domain is the one asset that must move with you.

Test every form with a real submission. AI builders love generating contact forms that look finished and route nowhere. Submit one yourself. Check the inbox. Check spam. Set up a second notification path if the business depends on leads.

Open it on two real phones, not the preview. The builder’s mobile preview lies by omission — it shows layout, not load time on a throttled connection, and not how the tap targets feel under an actual thumb.

Check the speed number, not the vibe. A 6MB hero image the AI picked looks great and loads like it’s 2011. Compress it. Free tools do this in seconds.

Put analytics on before launch, not after. You can’t learn what the first hundred visitors did if you install tracking on day thirty.

Write the legal pages yourself. Privacy policy, terms, refund or cancellation policy — the AI’s version is generic filler that may not match how the business actually operates.

Confirm the exit. Know, before you invest months of content, how you’d export the site or at least the content if you outgrow the builder. Some platforms make this trivial. Some make it a hostage negotiation.

None of this is hard. All of it is the kind of thing a non-technical owner reasonably assumes the AI already handled — because the AI presented it as handled. That assumption is the gap where the human fee still lives.

The Product This Actually Kills

Here’s our strong opinion: the thing dying isn’t “web development.” It’s a specific product — the $1,500 to $3,000 five-page static site, sold with no strategy, no copywriting help, no maintenance, and no accountability for whether the site ever produces a customer.

That product deserved to die. It was a template with the logo swapped, and every business owner who paid for one and got silence in return already knew it. The Freelancer.com numbers — 59% fewer web design ads, 58% fewer WordPress gigs — are mostly that product evaporating. Calling it good riddance is a defensible reaction.

But notice what didn’t collapse as hard in the same data: the complex work never lived on Freelancer.com’s job boards in the first place. Nobody posts “rebuild our booking infrastructure” as a $75 fixed-price gig. The developers getting hurt are the ones who sold the commodity. The ones who sold judgment — which platform, what structure, what the site should actually do for the business — are mostly fine, and some are busier than before, because AI-built sites keep generating the exact problems judgment solves.

If you’re a developer reading the 53% number with dread, the move isn’t to compete with the builder on its home turf. It’s the positioning play our solo developer playbook lays out: sell the outcome and the judgment, not the pages. “I build websites” is now a sentence an AI can finish. “I fix the reason your AI-built site isn’t generating calls” is a business.

One Honest Caveat

We should say the quiet part out loud: this survey was commissioned by Bluehost, a company that sells web hosting. A hosting company publishing research that says “building websites is easy now, come buy hosting” deserves at least one raised eyebrow. The sample is 201 owners, and the $850 figure is self-reported savings, not audited. TechRadar, which covered the findings, added its own caution on the Freelancer.com data: the job-board decline comes from a single marketplace and can’t by itself prove AI caused it. Some of that 59% is post-pandemic normalization; some is budgets moving to other platforms.

We’re not sure the AI did all of this. We’re fairly sure the commodity end of website work is permanently cheaper than it was, and that the direction of travel isn’t reversing. That’s enough to act on without pretending the data is cleaner than it is.

What a Small Business Should Actually Do This Week

Use the AI builder for version one. Seriously — for a brochure site, it’s the rational choice now, and the 53% aren’t wrong. But budget half a day of human attention for the checklist above, not zero. The 9% who shipped the first draft aren’t the role models here; they’re the survivorship bias. The 91% who edited are the actual playbook.

And keep the subscriptions honest while you’re at it — if the builder is one more monthly fee stacked on everything else, our breakdown of which AI apps actually earn their fee is the filter we’d run it through. Once the site is live, a simple brand kit is what keeps the website, the socials, and the ads looking like they came from the same business instead of three different ones.

The website isn’t the hard part anymore. Making it yours — owned, measured, and actually producing customers — still is. That part never was a template.

FAQ

Is the $850 and 32 hours saving real, or survey spin?

For a five-page brochure site, it’s real — that was genuinely overpriced work. It stops being real the moment the site needs to do something: booking with real availability rules, a quote calculator, checkout that doesn’t leak. And one honest caveat from the post itself: the figure is self-reported from 201 owners, in a survey commissioned by a hosting company.

Should I ship the AI’s first draft unchanged like the 9% did?

No — the post calls them survivorship bias, not role models. The 91% who edited before publishing are the actual playbook: budget half a day of human attention for the checklist, not zero.

What is the one thing I must own outright?

The domain, registered at a registrar you control — not inside the builder’s checkout. If the builder relationship sours, the domain is the asset that must move with you, and builders have been known to register it under their own account.

Does this mean freelance web development is dead?

The commodity end is: the $1,500 to $3,000 five-page static site with no strategy and no accountability. The judgment end — which platform, what the site should actually do for the business, why an AI-built site isn’t generating calls — is busier than before, because AI-built sites keep generating exactly those problems.