I build a product in this space, so read this with that in mind. I'll get to it at the end.
For about fifteen years a landing page builder was the sensible way for a non-developer to put a page on the internet. You picked a template, dragged blocks around, typed your copy into boxes, connected the form to Mailchimp and pressed publish. The editor was the product. Hosting, forms and a bit of analytics came bundled, so you never had to think about them.
The editor is the part that stopped making sense.
The page is the easy part now
If you can describe what you sell, ChatGPT or Claude will write you a decent landing page in a minute: a headline, a few sections, a form, CSS that holds up on a phone. Ask for three versions and pick one. Say "shorter, less corporate, move pricing up" and it does that too. That is the job the drag-and-drop editor existed for, and a chat does it faster, with nothing to learn.
What surprised me is that the assistants seem to agree.
On 15 October we asked eight AI tools (the ChatGPT app, OpenAI's API with search, Claude, Perplexity, Google's AI Overviews and AI Mode, Grok and Mistral) 68 questions people ask when they need a landing page. Things like "What are the best AI landing page builders in 2026?", "How do I host an HTML file that ChatGPT made?" and "How do I add a working contact form to an HTML page ChatGPT wrote?". None of them named a product. We sell to people asking the second and third kind, so the set leans that way. Keep that in mind.
Across about 1,560 answers, the two products named most often were not builders. They were hosts:
- Vercel: 492 of 1,561 answers (31.5%)
- Netlify: 480 of 1,538 (31.2%)
- Framer: 360 of 1,560 (23.1%)
- Unbounce: 352 of 1,537 (22.9%)
- Wix: 331 of 1,560 (21.2%)
- Leadpages: 258 of 1,560 (16.5%)
- Carrd: 213 of 1,536 (13.9%)
It's one day's run, and these answers change a lot from one run to the next, so don't read anything into the decimals. The claim I'd make is a small one: when you ask an assistant how to get a page live, it no longer sends you to an editor. It assumes you have the HTML and tells you where to put it.
(Our own product came up in 2 of 1,584 answers. Fair is fair.)
Where it actually breaks
The trouble is everything after the HTML. I read a lot of posts from people trying to do exactly this, and the same three problems come up every day.
Where does it go? The page is sitting in a chat window. Some assistants will share it, but the link lives inside their product and belongs to their account, not to you, and it's not something you'd print on a business card.
The form goes nowhere. An assistant will happily write a form with a submit button. Nothing receives it. People find out when the first lead never shows up, or they never find out. "My contact form doesn't send anything" might be the sentence I read most often.
Nobody knows what happened. A page you can't measure is a guess. Did anyone read past the headline? Which section do people leave at?
One post last week summed it up. Someone asked Claude for "something to collect emails for launch" and got a full todo app instead of the one-page waitlist they needed. Writing that page would have taken the model seconds. What it couldn't know was that a waitlist means a page plus somewhere for the emails to go.
Where builders still win
Builders aren't dead, and I'd be lying if I said they were. They still win when:
- several non-technical people edit the same pages every week and need a visual editor they all understand;
- an agency runs a brand system across dozens of clients, with approvals;
- you have enough traffic to A/B test properly, and the builder runs the test for you;
- you need a big catalog of integrations someone else maintains;
- you don't use an AI assistant at all, which is still most people.
- If that's you, keep your builder. This is about the growing group who already have the assistant open and pay for a builder only because it came with hosting and a form.
What replaces it
If the assistant writes the page, the rest of the stack gets small:
- somewhere to publish it under a real URL, with versions so you can undo a bad change;
- a form endpoint that delivers to an inbox or a CRM;
- measurement the same assistant can read, so it can say "people stop at the pricing section" and change one thing;
- a person approving what goes live.
You can watch this shape form. OpenAI put hosting into ChatGPT with Sites. Leadpages, a builder company, launched HTML Pub, whose pitch is roughly "your AI builds it, we publish it". Convika describes itself as a git repo, a static host, a form service, an analytics script and DNS, replaced by one MCP server. A new YC company, Spacesheep, launched on Product Hunt this weekend doing publishing for agents. Everyone is heading the same way: the editor goes, and what's left is plumbing an assistant can drive.
I'm building one of these, Pageree. It's an MCP server, so ChatGPT, Claude, Cursor and others can publish a page, send its form to your email or CRM and read section-by-section visitor data back, and nothing goes live until you approve it. It isn't the only way to do this. If you're a developer, a static host plus a form service plus an analytics script, glued together by hand, gets you most of the way.
The point I actually care about is narrower. If you're not a developer and you already write your page in a chat, paying for a visual editor you never open, just to get hosting and a form, doesn't make sense anymore.
Top comments (0)