What the AI web-design demos get wrong for a trades business

Open YouTube this week and you'll find a wall of videos with titles like "Your Web Designer is Obsolete" and "Does AI Kill the Web Design Industry?!" One creator's "$10K Website in Minutes" demo pulled 166,000 views in four days. Another channel's one-shot design tests cleared 640,000. A dozen of these videos shipped in the last three weeks, all built on the same new Anthropic model.

I watched most of them, because I build websites for excavators and supply yards with those exact tools. The demos are real. The model one-shots a scroll-animated 3D site that would have taken an agency a month, and it does it before your coffee cools.

Then I looked at what got demoed: a luxury watch brand, a designer portfolio, a fictional architecture studio, a perfume label. Across a dozen videos I counted zero plumbers, zero excavators, zero supply yards. None of these creators typed the prompt for a business where the website's whole job is to make a phone ring in a truck.

That gap matters more than the demos do. If you run a trades business and one of these videos left you wondering whether to fire your web guy or build your own site next weekend, this post is the missing second half.

The demos optimize for the wrong scoreboard

Every demo ends the same way: a slow scroll through a gorgeous page while the creator says "look at this." The scoreboard is beauty.

Your scoreboard is calls booked. A homeowner with a flooded basement gives your site about eight seconds on a phone, on cell service, standing in the flooded basement. She needs your number, proof you're licensed, and proof you serve her town. A cinematic 3D scroll sequence loads slower and answers none of that. The 4K video background that made the demo pop is the exact thing that pushes your load time past the point where she gives up and calls the next result.

A trades site that earns its keep looks boring in a demo reel: phone number pinned to the top of every page, one sentence saying what you do and where, license number, photos of your actual crew and your actual jobs, reviews, and a form that's three fields instead of nine. None of that gets views on YouTube. All of it gets you jobs.

The video ends when the session ends

Watch what happens at the end of any of these demos: nothing. The site exists on the creator's laptop. Nobody connects a domain, sets up redirects from the old site, or explains who updates the phone number in October when you add a second truck.

A real business site changes constantly. You add a service, drop a service area, raise a price, win a review worth featuring, lose a crew member whose face is on the About page. With an agency you emailed a guy. With a site you one-shotted from a YouTube tutorial, you are the guy. The AI session that built it is gone, and rebuilding the context takes longer than the original build did.

The form goes nowhere

Every demo includes a contact form. No demo shows what happens when someone submits it.

At a trades business the answer decides whether the site makes money. A lead that gets a response within five minutes books at several times the rate of one that waits a day. The homeowner filling out your form at 7 PM is filling out two competitors' forms too, and the first callback usually wins the job. A beautiful site wired to an inbox you check at lunch is a brochure.

The plumbing behind the page is the actual product: the form that texts the customer back instantly, the missed call that triggers an automatic text, the lead that lands in a system that nags you until someone calls. We build that layer for every client, and no part of it appears in any of the dozen videos.

Google can't see a pretty page

The demos never mention search. For a trades business, search is the site's day job: "excavating contractor Hebron KY" at 6 AM is worth more than every design award combined.

AI-generated sites routinely ship without the unglamorous parts that make local search work: LocalBusiness structured data, a separate page per service with its own title, a sitemap, consistent name-address-phone matching your Google Business Profile. The model writes this correctly when you ask. The videos never ask, because structured data doesn't demo well.

Nobody answers the money question

A few creators do address the business side, and their advice converges on "keep charging thousands, sell strategy instead of hours." Fair enough. None of them show what that means when a customer asks why a site that took 20 minutes costs $8,000.

The honest answer: the build was never the expensive part, and now everyone gets to see that plainly. You're paying for someone to know which twelve pages a septic company needs and which forty it doesn't, what a lead is worth in your market, what happens after the form submit, and who picks up when something breaks in month seven. The typing got cheap. The judgment didn't.

What this means if you run a trades business

Three takeaways, in order of money:

• Don't pay 2023 prices for a build. If an agency quotes you $15K and six weeks for a brochure site, they're billing you for labor that no longer exists. The tools got 10x faster and your quote should reflect some of that.
• Don't buy the demo. If a pitch leads with animations and mockups, ask instead: what happens when a lead comes in at 7 PM? Where do form fills go? Who updates the site in six months? What shows up when someone searches my trade plus my town? Whoever answers those four questions well is worth hiring, whether they use AI or not.
• The site is one piece of a system. We build ours on the same stack every time: the site up front, lead capture wired to instant text-back behind it, everything logged where the owner can see it. Our supply-yard and excavation clients get the boring-looking version, on purpose.

What I'm not claiming

• That the demos are fake. They're real, and the model is a genuine leap. I use it daily and it does the work of a small design team.
• That you need to spend big. A solo operator with a $500 site that loads fast and shows a phone number beats a $20K site that hides it. Spend on the response system before the redesign.
• That web designers are finished. The ones who sold typing are in trouble. The ones who sell judgment about what a specific business needs just got faster and cheaper to hire.

The actual decision

You probably don't have a website problem. You have a "what happens to a lead" problem, and the website is the most visible piece of it.

Two ways to find out which one you have:

1. Fill out your own contact form tonight at 7 PM and time how long until anyone responds. That number tells you more than any redesign pitch.
2. If the number embarrasses you, look at how our audit works or book a 15-minute fit call. We'll tell you honestly whether you need a new site, a response system behind the one you have, or neither.