pillar articles

Rented Ground: What You Actually Own When You Own a Website

Every major website platform is being rebuilt around code and AI agents. Four questions decide whether that’s your problem or someone else’s.

Something unusual happened in August

On 4 August, Webflow’s Chief Product Officer and a VP of Engineering published a piece admitting that when you design in Webflow, you aren’t editing your website. You’re editing an internal representation of it, which a compiler then turns into the thing that actually ships. Two versions of the truth, joined by a translation layer.

Their words, not mine: any system built that way will drift.

I’ve made a living on that platform for seven years and I’ve defended that architecture more than once. I think the admission is correct, I think publishing it took some nerve, and I think most people reading about it are drawing the wrong conclusion.

Because this isn’t a Webflow story. Framer is heading the same way. So are the AI code-generation tools, from the opposite direction. The entire category is converging on the same answer at the same time, and the reason is simple enough to say in one sentence.

Why every platform is arriving at the same place

AI agents read and write code.

If the real version of your website is a proprietary format that only one company’s compiler can read, then an agent working on your site is working on a translation. Every instruction has to round-trip through something it can’t see. So if you want agents to be genuinely useful — and every platform now does — the code has to become the actual artefact rather than an export.

That’s it. That’s the whole shift. Everything else follows from it.

What this means for you this month: almost nothing

Webflow was explicit that this was an announcement of a direction, not a release. They called it the hardest engineering work the company has taken on, and said they’re building the bridge before asking anyone to cross it. Their conference runs 1–3 September and we’ll all know more afterwards.

Here’s the planning assumption I’d use, and I’d use it for any platform making a similar promise: feature removals ship on schedule, platform rebuilds don’t. Webflow retired Logic in June 2025 and User Accounts in January 2026, both on the dates announced. A ground-up rearchitecture of a platform millions of sites depend on is a different order of problem. Plan in years.

Your site isn’t going to break. Nothing is being taken away from you next week. If you’re reading this feeling like you need to do something urgently, you almost certainly don’t.

But there are four questions worth answering properly, and they’re worth answering whether or not any of this ever ships. They’re the questions I now ask before recommending anything to anyone.

The four questions

1. What does it cost to run for five years?

Not what it costs to build. Building has got dramatically cheaper over the last eighteen months and is still getting cheaper. Running a website hasn’t moved, and in a couple of places it’s got more expensive — platforms have started metering AI usage in units they don’t fully explain.

Most owners negotiate hard on the number that stopped mattering and never calculate the one that did.

Building Got Cheap. Owning Didn’t.

2. Could you leave?

Not should you — could you. If you decided tomorrow to move your site somewhere else, what would you actually be able to take with you, and what would you have to rebuild from nothing?

Almost nobody knows their answer. It takes about thirty minutes to find out, and knowing it changes how you feel about everything else on this page.

The Exit Test

3. Is a visual builder still the right shape for you?

There are three broad options now — a managed visual platform, a headless CMS with a custom front end, or something fully bespoke — and the honest answer depends far more on who edits your site each week than on any technical argument.

I’ve written the comparison without a predetermined winner, which means every option loses at something.

Builder, Headless, or Bespoke

4. What happens when the platform removes something you depend on?

The realistic risk to your website isn’t your platform going under. It’s one feature your business runs on being handed to a third party with a date attached. It’s happened twice on Webflow in eighteen months, it happens constantly in the WordPress plugin ecosystem, and Google does it to its own products regularly.

There’s a way to work out in advance which of your dependencies are exposed.

When the Platform Removes Something You Depend On

Two more, once you’ve done those

Two questions that used to be someone else’s job and are quietly becoming yours.

Who’s actually allowed to change your website now? AI agents can edit production sites. That turns “who can publish” from an afterthought into a real governance question, even for a business of ten people.
Who Is Allowed to Change Your Website Now?

What do AI assistants tell people about your business? You have a second audience now, and it isn’t reading your site the way a person does. Increasingly it’s answering questions about you before anyone reaches you.
Your Second Audience

The rule I’d keep

If you take one thing from this: never rebuild a working website because of an announcement. Rebuild it because of a number.

Announcements are free to make and cost you tens of thousands to react to. Numbers — a five-year run cost you can’t justify, an exit cost you can’t stomach, a ceiling you’ve actually hit — are the only honest reason to move.

Where to start

Start with the exit test. It costs nothing, takes half an hour, and you’ll finish it knowing something concrete about your own position rather than something general about the industry.

If your score is fine, you can stop reading, get on with your week, and revisit this next summer. That’s a legitimate outcome and it’s the one most people should have.

Take the exit test

Prices, dates and platform features in this series were verified on 13 August 2026. This is a fast-moving area — check anything you’re about to act on.

Mark Thurman