Why We Replaced Webflow with Astro
For years, Webflow was the sensible answer when a client wanted a beautiful marketing site without the weight of WordPress. It isn't any more, at least not for us. We now build those sites on Astro, with AI doing much of the build and our developers leading and reviewing every change. This is why we made the switch, what it changes technically and, most importantly, what it changes for the marketing team that has to make the website earn its keep.

The short version
We moved our marketing-site work from Webflow to Astro for four reasons:
- Cost. Webflow bills per site and per seat in US dollars. At South African exchange rates, that’s an open-ended cost that rises without anything getting better.
- Ownership. A Webflow site can only really live on Webflow. An Astro site is a Git repository that belongs to the client and runs on any host.
- Limits. Webflow’s CMS and builder impose plan-based ceilings and push anything unusual into workarounds. Astro has no platform ceiling, only the limits of good engineering.
- AI. The biggest change in how websites get built in the last two years is AI that writes, tests and refactors code. It works best on plain code in a repository. A visual canvas leaves most of that advantage on the table.
The fourth reason is what turned a slow drift into a decision. The rest of this article explains each one, including what we gave up.
What Webflow got right
It would be lazy to pretend Webflow was a bad product. It wasn’t, and for many teams it still isn’t.
Webflow solved a real problem. Before it, a designer who wanted pixel control had two choices: hand a Figma file to a developer and wait, or fight a WordPress page builder that produced bloated markup. Webflow gave designers a visual canvas that output clean, standards-based HTML and CSS, with hosting, a CMS, forms and interactions in one subscription. For a marketing team with no developer, that’s a genuinely good deal.
But our situation is different. We’re a development studio. We’ve built websites since 2013, and we write code for a living. The things Webflow abstracts away (markup, styling, hosting and deployment) are things we’re good at and want control over. For us, the abstraction had stopped saving time and started costing it.
Reason one: the dollar bill that never stops
Webflow pricing has two layers. Site plans are billed per website and unlock hosting, CMS capacity and features. Workspace plans are billed per seat for the people who build and edit. Both are priced in US dollars, and add-ons such as localisation sit on top.
For a business in the US, that’s a manageable line item. For a South African business, it’s a recurring cost in a currency it doesn’t earn in. When the rand weakens, the website gets more expensive without improving. Multiply that by every client site on an agency’s books, and by the seats needed to build them, and it adds up to a significant amount of money each year spent purely on the right to keep the site online.
An Astro site builds to static files. Static hosting is among the cheapest infrastructure there is: Cloudflare, GitHub Pages and AWS can serve a marketing site for very little, and we can host on infrastructure we already run, billed in rand. The money a client used to send to a platform can go into content, campaigns or improvements to the site itself.
That isn’t an argument that Astro is free. Someone still builds the site, and someone still maintains it. But the cost is now for work, not for permission, and it’s in a currency the client earns.
Reason two: you should own what you paid for
This is the reason clients understand fastest once it’s explained.
Webflow offers a code export, and it’s tempting to read that as ownership. It isn’t. The export gives you static HTML, CSS and JavaScript for your pages. What it doesn’t give you, in working form, is the stuff that makes a site a site: the CMS (collection content comes out separately, as data files, not as working templates), forms and e-commerce. Interactions depend on Webflow’s own runtime. In practice, leaving Webflow means rebuilding.
Lock-in matters most when you need to leave and can’t. That can happen because:
- pricing changes and the new plan no longer fits;
- a feature you need isn’t on the roadmap;
- a new marketing director wants the site on the company’s own infrastructure;
- the company is acquired, and IT wants every web property on one standard.
An Astro site is a folder of source code in a Git repository. The client owns the repository. It builds with open-source tools under a permissive licence, and it deploys to any static host. If they part ways with us, another competent developer can clone it and keep going the same afternoon. We’d rather keep clients because they want to stay than because they can’t leave.
Reason three: limits you only hit once you’ve grown
Webflow’s CMS is pleasant for simple content: a blog, a team page, a list of case studies. The friction comes later, usually just as the site starts to matter.
Capacity ceilings. CMS items, collections and fields per collection are all capped by plan. A product catalogue, a location directory or a resources library can run into those ceilings, and the fix is a bigger plan in dollars, not better engineering.
Rigid content structures. Relationships between collections are limited, and complex content (a product with variants, specifications, compatible models and downloadable documents) has to be bent into the shapes the CMS supports.
Logic by workaround. Anything the builder doesn’t do natively, such as a calculator, a filter with real logic, a pre-filled enquiry form or a CRM integration, means custom embed code pasted into the canvas, third-party scripts or an automation service in the middle. Each one works, but together they become a fragile layer nobody fully owns.
In Astro, content is data in a content collection with a schema. A product’s fields, types and relationships are defined in code, and the build checks every entry against the schema. A missing image, a broken reference or a price in the wrong format fails the build on our machine instead of appearing on the live site. There’s no item ceiling to buy your way past, and when a page needs real logic, we write it properly as a component, not as a snippet in an embed box.
Nero Studio is a good example. It’s a premium automotive accessories brand with a catalogue that started life as a 24-page PDF. In Astro, that catalogue became a typed product data layer that drives a filterable catalogue and an enquiry flow that pre-fills the product the visitor was looking at. That’s routine in a codebase. In a visual builder, it’s the kind of thing that turns into a project.
Reason four: AI works on code, not canvases
This is the reason that made the switch urgent.
AI coding tools have changed what a small, experienced team can deliver. We use Claude Code in our build process. Working inside a repository, it can build components from a design, wire up content, write and run browser tests, fix what the tests catch and refactor without breaking things. On the Nero Studio build, that took the site from catalogue PDF to a tested, animated website in a few hours rather than the weeks it would traditionally take.
All of that depends on the work being plain code under version control. AI is most reliable when:
- the whole project is readable as text, so it can understand how the pieces connect;
- every change is a diff a human can review before it’s accepted;
- there are tests and a build step to catch mistakes immediately;
- a bad change can be reverted in seconds.
A visual builder stores the source of truth in a proprietary project format behind someone else’s interface. Even with AI features bolted on, there’s no repository to reason over, no diff to review, no test suite to run and no clean way to roll back one change. Using AI on a canvas is like having a fast junior developer you can’t supervise.
Astro is the opposite. It’s an ordinary web project: components, content files and configuration, all in Git. AI can do most of the typing, and we can see and judge every line.
AI-assisted, but human-led
“AI-built website” makes some people nervous, and they’re right to be cautious. Unsupervised AI produces sites that look plausible and fall apart under scrutiny. So the line matters, and ours is clear.
People own the decisions. Positioning, information architecture, the content model, design in Figma, technical choices and what ships are all decided by our team. AI doesn’t decide what a page is for.
AI does the production work. Components, data wiring, responsive images, tests and the repetitive work across dozens of templates.
Every change is reviewed. It arrives as a commit. We read the diff, run the tests and check the result in a real browser. If we can’t explain a line of code, it doesn’t go live.
The test we hold ourselves to is simple: after the build, we still understand the codebase. Speed that leaves you with a site nobody can confidently change isn’t speed at all. It’s just a debt that hasn’t come due.
What Astro actually does differently
For the technical readers, here’s what we get from Astro specifically, rather than from “using code” in general.
HTML first, JavaScript only where needed. Astro renders pages to HTML at build time and, by default, sends no JavaScript to the browser. Interactive parts such as a filter, a calculator or a carousel are islands: individual components that load their own JavaScript, optionally only once they’re visible. The rest of the page stays static HTML. Most marketing pages are mostly content, so most of the page costs the visitor nothing to run.
Use any UI framework, or none. Islands can be written in React, Vue, Svelte or Solid, or as plain Astro components. We can bring in a mature component where it helps without committing the whole site to a single-page-app framework.
Content collections with schemas. Content lives in Markdown, MDX, JSON or YAML, or is pulled from a CMS or API, and is validated against a schema at build time. Editors get guardrails, and developers get type-safe content in their components.
Built-in image optimisation. Images are resized and converted to modern formats such as AVIF and WebP at build time, with correct dimensions to prevent layout shift.
Deploy anywhere. Static output runs on any host or CDN. When a page genuinely needs server rendering, adapters support that too, so the site isn’t trapped as a purely static site.
Small attack surface. A static site has no admin login on the public site, no database to inject into and no plugins to patch every week. For our clients, that means fewer late-night security updates and less risk.
What it means for the marketing team
This is where the switch pays off, and it’s the reason we’d recommend it to most marketing-led businesses.
Speed is a marketing metric
Page speed isn’t just a technical vanity number. Core Web Vitals feed into how Google evaluates page experience. In paid search, landing-page experience is part of how ads are assessed, and a slow page wastes the budget you spent to get someone there. We manage about R5 million a month in ad spend, so we see the cost of a slow landing page directly. Astro’s HTML-first approach makes good Core Web Vitals the default rather than something you fight a platform to achieve.
Landing pages at campaign speed
Campaigns need pages quickly: a variant for each audience, each offer and each market. On Astro, a landing page is a set of reusable components with typed content. With AI in the workflow, a new variant, a new section type or a new localised version is a quick, reviewed change rather than a manual rebuild on a canvas. More variants, tested sooner, is how paid media improves.
Technical SEO you control
Every meta tag, canonical URL, redirect, structured-data block and sitemap entry is in code, so it’s generated consistently and reviewed like everything else. Schema markup for products, services, articles and FAQs is built into the templates rather than pasted page by page. Clean HTML with no runtime framework is also easy for search engines and AI assistants to read.
Tracking without the tangle
Analytics, tag management and conversion events are part of the codebase: GA4, Google Tag Manager or a privacy-first tool such as Umami. Events are named consistently, forms report properly, and a tracking change is reviewed and versioned instead of being a script someone pasted into a site setting two years ago.
Content that’s safe to edit
Marketing teams still need to change content without calling a developer, so we set up the editing workflow per site:
- A Git-based CMS (our preference). Tools such as Decap CMS and TinaCMS give editors a friendly interface, and every save becomes a commit. You get full history, previews and one-click rollback, and the content lives with the code.
- A headless CMS. Sanity or Storyblok for teams that want a hosted editor, structured content or visual page blocks.
- Headless WordPress. For teams who already know WordPress: edit in the dashboard, and our pipeline publishes a static Astro site.
- Managed changes. On many brochure sites, the simplest option is to send us the change.
Editors work within structures designed for the content, so they can change copy, images and entries but can’t accidentally break a carefully designed layout. That’s the same view we take on page builders in WordPress: give people the power they need, not the power to break things.
Costs you can plan
Hosting in rand, no per-seat platform licences on a Git-based setup and no plan upgrades triggered by content growth. The website budget goes on the website.
What we gave up
An honest comparison has a cost column. These are the trade-offs we accepted.
Designer self-service. In Webflow, a designer can change a layout and publish it without a developer. In Astro, a layout change is a code change. With AI in the loop, it’s a fast one, but it still goes through us. For teams with in-house designers who love Webflow and don’t mind the bill, Webflow can still be the right tool.
The visual interactions editor. Webflow’s interaction timeline is a pleasant way to build animation. In Astro, motion is written in code with CSS and lightweight libraries. The result is usually lighter and respects reduced-motion settings by default, but it’s not a drag-and-drop experience.
One subscription for everything. Webflow bundles hosting, CMS, forms and an editor. An Astro site is assembled from parts: a host, an editing setup, a form handler and analytics. We choose and configure those parts, and we document them, but there are more of them.
Built-in commerce. For a full online store, we’d still usually recommend WooCommerce or a dedicated e-commerce build. Astro handles catalogues and enquiry flows very well, but it isn’t a store in a box.
For the sites we build (marketing sites, campaign pages, catalogues and content-led brand sites), those trade-offs are clearly worth it.
When Webflow is still the right call
We’d still point some teams to Webflow:
- an in-house design team that wants to own layout changes, with no developer involved;
- a business billed in dollars, for whom the subscription is immaterial;
- a short-lived microsite where lock-in genuinely doesn’t matter.
If none of those describes you, and the site is meant to perform for years, Astro is the stronger foundation.
Moving a site from Webflow to Astro
If you’re on Webflow and the reasons above sound familiar, here’s how we move a site:
- Audit. Pages, CMS collections, forms, interactions, integrations, tracking and every URL that has search rankings or ad traffic.
- Content model. Webflow collections become typed Astro content collections, usually with better relationships than the original.
- Rebuild. The design is rebuilt faithfully, or improved if you want that, as reusable components, with AI doing much of the production work and our team reviewing it.
- Content migration. Collection content is imported from Webflow’s data exports and checked against the new schemas, so anything missing or malformed shows up before launch, not after.
- SEO continuity. URLs are preserved or redirected one to one, and metadata, structured data and sitemaps are carried across.
- Editing setup and training. Git-based CMS, headless CMS or headless WordPress, with a short guide for your team.
- Launch and support. Testing on real devices, hosting, monitoring and a 24-hour support response time.
The bottom line
Webflow made beautiful websites accessible to people who didn’t write code. We do write code, and with AI the cost of writing it has dropped sharply, while the value of owning it has stayed the same. On Astro, our clients get faster sites, content structures that grow with them, hosting in rand and a repository they own. They also get a team that understands every line, because we read every line.
See the Astro development service for how we build these sites, or read the Nero Studio build for a step-by-step look at the AI-assisted workflow.
Questions about Webflow and Astro
A few useful answers before you decide.
Is Astro better than Webflow?
Can I edit an Astro site without knowing code?
Will moving from Webflow hurt our SEO?
Is an AI-assisted website lower quality?
What does an Astro site cost to host?
Keep the right things connected.
Astro development: fast websites you own, built with AI and led by people
Astro is what we build marketing websites on now. AI does much of the typing, we make every decision and review every change, and you own the code, the content and the hosting.

ChatGPT as Creative Director, Claude as Build Team: A Website in Hours
How we took an automotive accessories brand from a catalogue PDF to a tested, animated Astro website in a few hours, and why we still understand every line of the code.
Nero Studio — Premium automotive accessories website
Brand positioning, AI-generated imagery and an Astro build for a premium automotive accessories brand, taken from a catalogue PDF to a tested site in a few hours.

WordPress vs Custom PHP: Choosing the Right Business Website
Choose between WordPress and custom PHP by examining publishing needs, business workflows, integrations, maintenance and long-term ownership.
Start a conversation
Tell us what you are working towards. We’ll help you find a practical next step.
