Built to win the meeting

Three years ago, near the end of my consulting run, I sat in on a pitch a design agency made to a client of ours, a regional retailer with a website it was embarrassed by. Halfway through the deck the designer closed it, opened a browser tab, and rebuilt the client's homepage in front of the room. Not a sketch of a page. A working page, with the client's actual copy in it, scrolling, and when he dragged a divider across the canvas the whole layout folded itself into a phone screen and kept working. The managing director leaned forward. The CFO, who had spent the previous meeting demanding a cheaper option, said "wow" in a flat voice, as if he regretted it on arrival. The agency won the work that week. The site they eventually shipped was fine, on schedule, and unremarkable. I've thought about that meeting a great deal since I started reviewing these tools, because the fifteen minutes were the sale and the four months were the product, and the gap between them is where Framer lives.

The tool the designer used was Framer. The claim I want to weigh is this: Framer is not a website builder with a good demo. The demo is the product. Everything else about it, the pricing, the hosting, the missing export button, follows from a company that has engineered the first fifteen minutes with a precision it has never applied to anything that happens after them.

The first fifteen minutes are a design target

What the tool actually does is two things stitched together. Type a prompt and the AI, called Wireframer, generates a responsive multi-page wireframe in under a minute, headlines already written, sections where you asked for sections. Then you drop into a canvas that behaves like Figma, where every pixel is editable without code, and the content lives in a built-in CMS that reads like a spreadsheet rather than like a database with opinions. Publishing is a click and a few seconds, hosting included. The free plan carries no AI credit limits, which sounds like generosity until you notice what the free plan is for. It is a demo room you can carry into somebody else's meeting.

The hands-on reviews agree on what the machine does and does not do. Walter Akolo's review for HostAdvice found the generator got roughly 70 percent of the way to a finished site: mobile headlines that overflow the screen, a nested-link error that would break in some browsers, dropdown fields requested in the prompt and delivered as plain text boxes (the review is here). That last 30 percent is the actual job, and the tool gives you good instruments for it, which is a genuine compliment. But look at where the effort went. Consider the incentive: nobody at Framer is paid for the four months after the pitch. The tool is judged, bought, and posted about in the fifteen minutes, and a designer who wins a client mid-meeting becomes an acquisition channel wearing a lanyard. The product has been tuned, rationally, to the moment it gets discussed.

The export button that is not there

Now the mechanism underneath the rent. Framer does not export your site. No HTML, no CSS, at any tier, and the thing you build lives on Framer's servers for as long as you keep paying, which is a fact the reviews keep listing in the cons column and the marketing keeps leaving off the page. The pricing makes the structure legible:

The add-ons tell you where the curve goes: twenty dollars per translation locale, fifty dollars per half million A/B testing events. Your bill climbs with your success, which is the standard shape for a hosted platform and not, by itself, a scandal. What is not standard is what leaving costs. On a platform that exports code, departure is a migration, a weekend and a freelancer. Here, departure is a rebuild, and the difference between those two words is the entire risk. In my last year of consulting I sat on a steering committee for a company that had built its whole marketing presence on a platform whose pricing committee it did not sit on. The renewal arrived 40 percent higher. The rebuild estimate was $180,000. They renewed, and the meeting took eleven minutes. I think about that arithmetic whenever a pricing page advertises how easy it is to start, because starting is priced as a temptation and leaving is priced as a project, and only one of those numbers is on the page.

The case for renting the stage

The counterargument deserves to be made better than its own proponents usually make it. A marketing site is a costume, and costumes get replaced. Companies redesign every two or three years; a landing page built for one campaign or one funding cycle will be retired before any platform's terms could matter to it. Meanwhile the alternative to renting is not free. A self-hosted site is a small facility that needs its plugins patched, its backups tested, and a freelancer on retainer, and those costs arrive monthly whether the site earns or not. At ten dollars a month, Framer bundles hosting, a CDN, a CMS, and a design tool, and the reviewer's published test site loaded fast, which is the part of the deal that actually touches customers. For a person whose site is a costume, the lock-in is a theoretical harm scheduled comfortably after the site's own retirement. I hold this argument honestly, and I'll note that my own incentives aren't clean: I publish on a site that handles the infrastructure, own no server, and pay no hosting, so my lectures about ownership are cheaper than they sound.

The site that has to outlive the trend

It depends, and the dependency is whether the site is a costume or an archive. A campaign, a portfolio you plan to rework, a product page with a season on it: rent the stage, enjoy the fifteen minutes, and budget the hours to fix the last 30 percent, because the machine will not. A site that accumulates, essays, documentation, a decade of posts, is a different object, and the export button you never miss is precisely the one whose absence you discover when the archive is too large to rebuild from a template. Watch one variable over the next couple of years: the free plan. The unlimited AI generation is the demo room's furniture, and AI costs are real costs; if Framer starts metering it, that tells you the company has stopped buying growth and started collecting on it. The demo was never lying. It just wasn't the invoice.