You’ve seen the video. Some guy with a ring light and a Claude tab open tells you the move: find a local business with a website that looks like it was built in 2009, cold-email them a screenshot next to a slicker AI-generated version, and watch them sign on the spot because you just handed them a “$10,000 website” for the cost of a prompt and an afternoon. It’s become one of the single most repeated plays in the entire AI-creator space — automation guys, agency bros, whoever just discovered Claude Code last month, all of them, at some point, doing some version of this same pitch.
I want to take this apart, because there’s a real idea buried in there and also a genuinely predatory practice sitting right on top of it, and most of the people making this pitch either can’t tell the two apart or don’t want to.
The Number Is Made Up
Let’s deal with the easy part first. The “$10,000 website” framing is itself kind of bullshit, and not in a subtle way. Go look at what a real high-converting site actually costs when a business with money commissions one — the kind of site you’d expect from Nike, Samsung, Apple, or whichever tech company is having a moment. Complex scroll states. 3D elements. Video-driven interaction design. Those sites aren’t $10,000. They’re $40,000, $50,000, sometimes six figures, and nobody who’s ever actually built one for a paying client would tell you otherwise.
I know designers personally who can produce something that reads, feels, and converts like a $5,000-effort site and sell it for $40,000 without blinking. That’s not a scam. That’s pricing based on outcome and perceived value, which is how design has worked since forever. The point isn’t that “$10,000” is a lie because the number’s too high. It’s a lie because it’s too low for what’s actually being implied — a number chosen specifically because it’s believable to a small business owner, not because it reflects what real work in that tier costs. It’s a marketing number, not a production number.
That’s the smaller problem, though. The bigger one is what’s not being sold at all.
The Looks/Works Fallacy
Here’s the thing nobody selling these websites wants to say out loud: when a client buys a website, they are not buying one thing. They’re buying two things, stapled together, whether they realize it or not. They’re buying how the thing looks — the customer-facing experience, the part that makes a stranger stop scrolling and actually read the page. And they’re buying how the thing works — whether anyone on their own team can open it up afterward and add a new employee’s contact info, update a price, swap out a photo, without calling someone.
The entire “$10,000 AI website” pitch, as it’s currently practiced, sells exactly one of those two things. It’s a before-and-after screenshot. Ugly site, pretty site, close the deal. Nobody in that pitch is talking about what happens six weeks later when the business needs to make a change nobody thought to mention during the sales call.
This isn’t a small oversight. It goes against one of the actual, foundational truths of why design has business value in the first place: design inspires use. Not just customer use — internal use. A business that’s proud of its website keeps it updated. They treat it like an asset instead of a liability. But that only happens if the people inside the business can actually touch the thing. A site nobody in-house can operate doesn’t get maintained, doesn’t get loved, doesn’t get used as “everything” the way a business owner wants their site to function. It gets frozen the day it launches and starts rotting the day after that.
Selling the look without the operability isn’t selling half a website. It’s selling a photograph of a website and calling it the real thing.
The Dependency Trap
Here’s where this stops being a marketing critique and becomes something closer to an ethics problem. Picture who’s actually buying these sites: not tech companies, not agencies, not people who’d ever crack open an HTML file for fun. A lawyer. An HVAC repair technician. A roofing contractor. People running real, profitable, non-technical businesses who are being sold a beautiful site by someone who, frankly, thinks their old one looks like shit — which, fine, it might.
Now fast-forward six weeks. That HVAC guy just hired someone new and needs to add their name and number to the contact page. He can’t. He doesn’t know CSS. He doesn’t know PHP. He doesn’t know JavaScript. The only person who can touch that site is the arrogant twenty-six-year-old who cold-emailed him in the first place, and that twenty-six-year-old is now the only path back into his own business’s front door.
That’s not a website sale. That’s a hostage situation with better branding.
I want to be precise about what I’m actually claiming here, because it matters. I am not saying every person selling a $10,000 AI-generated website is a con artist. Some of them are probably being completely honest about their process and their results. What I’m saying is that if the site leaves the client without the means to operate it themselves, the seller has built dependency, not value — and it doesn’t matter how good the launch-day screenshot looked. You can create a relationship where the customer has no real alternative except to keep paying you for every future change, because you never actually equipped them to make one. Some people will tell you that’s just smart business — leverage the knowledge gap, that’s the game. Sure, fine, to a point. But the point where it stops being smart business and starts being predatory is exactly the point where the client is left in a worse operational position than the ugly-but-functional site they had before you showed up.
The Locked Hood Problem
Think about buying a used car from a lot where every single vehicle has been detailed within an inch of its life. Paint’s flawless, interior smells like a new-car candle, tires are blacked out and glistening. You drive it off the lot feeling like you won. Three weeks later the check-engine light comes on, you pop the hood, and it’s welded shut. Not because the dealer did anything to sabotage you specifically — the car was never built with the expectation that you’d need to get in there yourself. It was built to sell, once, at the moment of the test drive. Everything after that moment was somebody else’s problem, and now that somebody else is you, calling the dealership back and hoping they still pick up.
That’s the $10,000 AI website. The paint job is real. The detailing is real. Nobody’s arguing the site doesn’t look good when it launches. What’s missing is the hood release — the basic, expected ability to get in there yourself when something needs to change, which for a website is not an if, it’s a when. Every business’s contact info changes eventually. Every business hires someone new eventually. A website that can’t absorb that without a phone call to the person who sold it to the business isn’t a finished product. It’s a beautifully finished demo with the maintenance door welded shut, and the client doesn’t find out until they need to open it.
Why This Is Predatory Right Now — But Not Forever
I want to be careful here, because this isn’t an argument that the $10,000 AI website is inherently a scam as a category. It’s an argument that the way the tooling is currently configured makes that specific pitch predatory right now, and that’s a meaningfully different claim.
The insinuation baked into the pitch is: “you get a beautiful, high-converting site AND you keep real control over it.” That combination — high creative output plus genuine post-handoff editability — isn’t something that currently exists in any tool I’m aware of. Not yet. So when someone sells that combination anyway, they’re selling something that doesn’t exist. Whether they know it or not is almost beside the point. The client experiences the same outcome either way: locked out of their own asset.
This is a market-maturity problem, not a character problem. The tooling hasn’t caught up to the pitch, and making the pitch anyway — ahead of the tooling actually being able to back it up — does real damage. It burns trust in AI-assisted design broadly, and it leaves actual small businesses stuck holding something they can’t operate.
The Tell
Here’s how I know this isn’t just a hunch. Right now, if you go looking for a tool that combines a genuinely high design ceiling — the kind of visual sophistication something like Claude Design or Claude Code can currently produce — with real, meaningful client-side editability after handoff, you will not find one. You get one side of that trade or the other. Tools that let you edit freely but cap what you can actually produce. Tools that let you produce almost anything but lock you into a rigid, low-flexibility editing experience once it’s built. Nobody has cracked the middle.
I know this because I actually watch the space where this kind of announcement would happen. I’m part of the Framer community online. I remember exactly what it looked like when Framer first broke out — the design world lost its mind over it, loudly, immediately, everywhere. If a tool existed right now that could genuinely carry the production burden through prompting while preserving real flexibility and control for the client after handoff — something that plugged into or rivaled Framer, Webflow, or MagicPath — we would be hearing about it with that same energy. We are not hearing about it. Not from Reddit, not from the professional Framer crowd, not from YouTubers, not from anywhere. The silence isn’t proof of nothing. The silence is the data point.
Maybe I’m wrong. Maybe something exists that I just haven’t run across yet, and I’ll happily eat that if it turns out to be true. But I’d wager, given how fast this specific corner of the internet reports on itself, that if the tool people are implicitly claiming to be using actually existed in the form they’re selling, it would already be a known name. It isn’t. Which means the pitch is running ahead of the product.
This is where I think a lot of the AI-content space fundamentally misjudges its own audience. The design community is not shy. It does not sit quietly on genuinely useful tools out of modesty or restraint. When something actually closes a real gap — Figma did it, Framer did it, even smaller utilities do it on a weekly basis — the response is immediate, loud, and impossible to miss, because designers have spent years getting burned by tools that promised flexibility and delivered rigidity, or promised power and delivered chaos. A tool that actually cracked the looks/works problem wouldn’t need a marketing push. It would announce itself.
Grow Into the Pitch Before You Make It
None of this means the $10,000 AI website is a permanently bad idea. If anything, closing this exact gap — real design ceiling paired with real post-handoff client control — is one of the more interesting problems sitting in front of anyone working seriously in AI-assisted design right now, and I know I’m not the only person quietly circling it. Some of that work is already happening, in various forms, in various places, mine included, and none of it is finished enough yet to talk about with a straight face.
But that’s exactly the point. The gap between the pitch and the actual tooling is the entire problem, and until that gap closes, selling the pitch is selling a photograph of a solution instead of the solution. The people making this pitch right now aren’t necessarily lying. They’re just early — early enough that being early is, functionally, indistinguishable from being wrong.
A beautiful website that nobody at the business can touch isn’t a $10,000 website. It’s a really nice-looking liability with a due date nobody’s told the client about yet.