Webflow vs Astro looks like a question about technology. It isn't. It's a question about who changes the layouts and how often, and that answer settles most of it before a single feature gets compared.

The short version
  • Stay on Webflow when a non-technical team changes layouts, not just words. The visual editor is the product, and it's what the seat-based pricing pays for.
  • Move to Astro when layout changes are rare and a developer, or an AI assistant working in the repo, makes them. You get very fast pages, full control of the page head and nothing to log into.
  • Webflow pushes back in the page head, not on the canvas. Its Data API has no canonical field, custom code goes through OAuth apps only, and GA4 can run first-party through your own domain.
  • Leaving costs more than the export button suggests. Code export is a paid Workspace feature that gives you markup; CMS content comes out separately, 100 items per API request.
  • Astro isn't a CMS. Decide the editing story first: Markdown in git, a headless CMS, or keeping Webflow for the pages marketing owns.
  • We've never shipped a Webflow site. The Webflow side of this post comes from auditing one, working in its API, and Webflow's own documentation.
0
Webflow sites we've built, and we say so below
0 of 28
pages with structured data on the Webflow site we audited
0
GA4 events from that site's main enquiry button
100
CMS items per request when you pull content out

Webflow vs Astro: what you're actually choosing

Webflow is a hosted visual builder. Design, CMS, hosting and publishing live in one place you log into, and as of November 2026 Webflow's pricing is a plan per site plus a Workspace plan for the team, with seat types that separate the people who design from the people who edit content and the people who only review.

Astro is an open-source web framework. Pages and components live in a code repository, Astro builds them into static HTML, and you host the result wherever you like. One is a product your marketing team uses. The other is a codebase your developer owns.

So the useful question isn't which is better. It's the row from our platform comparison that nobody weighs at the start: who adds a new page type? On Webflow, it's often the marketer. On Astro, it's a developer. If your marketing team builds a new landing-page layout every fortnight, Webflow is earning its keep and this post should mostly reassure you. If the layouts haven't changed since launch and every edit goes through an agency anyway, you're paying for a visual editor nobody uses.

Where we're standing

We should say this plainly: we have never built a Webflow site. There's no Webflow build in our portfolio, which is why we don't sell Webflow development as a service. What we have done is audit a Webflow site for a coworking operator, work inside Webflow's API on that same site, and scope and prototype its move to a custom static build. The Webflow side of this post comes from that work and from Webflow's documentation, read in November 2026.

Our view of the static side comes from building static sites. Our own site is static, though not Astro: plain HTML plus a Markdown blog that a small Python script builds in CI, with merging to main as the deploy. Lamacraft, a Calgary deck builder, came to us on WordPress with a brief to move to a static build, and the plan named Astro. The scroll-driven build sequence we made for them, one parametric spec driving both a real-time WebGL scene and a Blender render, is the kind of feature that's most comfortable when you own the whole build.

That's the vantage point. Read the Webflow column as an informed outsider's, and weigh it accordingly.

Webflow, honestly

Right when: a marketing team needs to build and rearrange pages without filing tickets, and you're happy being a tenant. That's the hub's default and we're not going to argue with it. A marketer who can rebuild a section on a Tuesday changes how much a company gets out of its website.

Wrong when: the layouts are effectively frozen and a developer makes every change anyway, or the site's job depends on things that live in the page head (canonicals, structured data, event tracking) and you want them generated rather than pasted. Also wrong when the site has to integrate deeply with another system, which is where the coworking operator below ended up.

What people underestimate: how much the platform carries. Hosting, a CDN, security patches, form handling, staging on a webflow.io subdomain, publishing. The site we audited was genuinely fast and clean, and none of that took any effort from its owner.

What people overestimate: the animation lock-in. Webflow now backs GSAP, and GSAP is free for all users, so the animation library itself isn't a reason to stay. What Webflow adds is a visual timeline for people who don't write code. If nobody on your team uses that timeline, the moat is smaller than it looks.

Astro, honestly

Right when: the site is mostly marketing pages, speed and calm matter more than in-house layout editing, and changes are occasional. Astro's islands architecture renders pages to static HTML and strips client-side JavaScript from components by default. You opt specific components back in with client:* directives, and server islands (server:defer) let a dynamic piece render separately while the rest of the page loads straight away. For a marketing site, that means the default is a page that ships no script at all.

Wrong when: non-technical people need to create new page types on demand and there's no CMS wired in. A static site without an editing story ossifies the day the developer leaves.

What people underestimate: that the editing story is a separate decision. Astro isn't a CMS. Its CMS guide lists dozens of headless options, WordPress, Sanity, Storyblok, Decap, Keystatic and TinaCMS among them, and Webflow isn't one of them. Forms, previews and hosting are separate decisions too.

What people overestimate: the risk. The company behind Astro joined Cloudflare on 16 January 2026, and its announcement says Astro stays open source and MIT-licensed and keeps supporting deployment targets other than Cloudflare's. For a marketing site, the code you own stays yours.

What the Webflow audit actually found

The coworking operator's Webflow site was in better shape than most sites we audit. All 28 pages returned cleanly, the median server response was about 0.15 seconds, and pages were around 10 KB of HTML. Fast, clean, well branded.

It was also under-built for search. None of the 28 pages carried structured data. Thirteen FAQ pages were near-empty stubs with generic titles. The location pages were thin and templated, there was no blog, and 115 of 116 images had empty alt text. Our recommendation was not to rebuild. Every one of those is fixable inside Webflow.

Measurement was the bigger gap, and it's the one worth learning from whatever platform you're on. GA4 was installed, but the site had no form element anywhere. Every lead path was a phone link, an email link, a chat widget, or a booking modal loaded in a cross-origin iframe from the coworking management platform. When we tested it in September 2026, clicking "Book a Tour" in a real browser fired no GA4 event at all and didn't change the URL. Tour requests were invisible to analytics by construction. That isn't Webflow's fault, and moving to Astro wouldn't fix it on its own; somebody has to build the bridge either way.

The owners decided to move anyway, and their reasons were real ones. They wanted the site to connect directly to the platform that runs their memberships and bookings. They wanted to edit copy and images themselves, and to make future changes by asking an AI assistant working in a terminal. Our recommendation was Astro, a GitHub repository, Netlify hosting and the platform's API, with payments and member logins staying on the platform. One lesson from scoping it: the live site's booking widgets came from a different platform than the one the integration was planned around. Before you wire a single form, get the client to say in writing which system is the system of record.

The page head is where Webflow pushes back

The canvas is Webflow's strength. The <head> is where it gets awkward, and for a site whose job is to be found, the head matters.

A Site API token can update a page's title, slug, SEO title and description, and Open Graph fields; Webflow's page settings endpoint documents exactly that, and no canonical URL field. Canonicals have to go in as per-page custom code. And the custom code endpoints require a bearer token from an OAuth flow, so a plain site token can't write them. Webflow's own route around that is its MCP server, which its documentation says can read and write site- and page-level custom code.

Three smaller things caught us out, all worth knowing if you're auditing or migrating one:

  • Token scope. A Workspace token generated for us carried only workspace-admin scopes, so a request to list sites came back 403 with a missing-scopes error. Webflow's token introspection endpoint returns a token's scopes with no side effects. Check first.
  • Attribute order. Webflow writes meta tags with content before name. A crawler regex that expects name="description" content="…" will report every description on the site as missing. Parse attributes; don't pattern-match their order.
  • First-party analytics. Webflow's GA4 integration can send hits through the site's own domain. On the site we audited, every event went to a path on its own host, and only an empty transport check went to Google. Faking or blocking that check stalls the tag, which looks exactly like "the tag doesn't fire". Worse, a test harness that blocks Google's hostnames lets real hits through. We learned that by leaking a batch of test page views and events into the client's live property. Block by request pattern, and answer every outgoing analytics request locally while you test.

None of this makes Webflow wrong. It does mean that if structured data, canonicals and event tracking are central to the site's job, you'll be working around the platform through OAuth apps and pasted code. On a static build they're just files: on our own site, the build writes the canonical URL, Open Graph tags and Article schema for every post.

Webflow vs Astro, side by side

Webflow Astro
Who changes layouts A marketer, in the Designer A developer, or an AI assistant in the repo
Who changes words Anyone with an editing seat Whoever the CMS or repo lets in
Pricing model Plan per site, plus Workspace plan and seats Free and MIT-licensed; you pay for hosting and any CMS
Hosting Webflow's Anywhere that serves files
Canonicals and schema Per-page custom code; API writes need OAuth Generated by the build
JavaScript shipped What the platform and your interactions need None by default; islands opt in
Animation Visual timeline, GSAP underneath GSAP or anything else, in code
Leaving Code export on paid Workspace plans; CMS via API Already files you own
Typical year-three failure A CMS or page-head limit you route around No editing story, so the site ossifies

The rows split cleanly. Webflow wins every row about people who don't write code. Astro wins every row about control and exit cost. That's why the decision is about your team, not the software.

Leaving Webflow: what comes with you

As of November 2026, code export sits on Webflow's paid Workspace plans. It gives you the markup Webflow generated, which is a useful reference and a poor foundation: you'll rebuild the layouts as Astro components rather than adopt Webflow's class structure. CMS content comes out separately. The Data API's list items endpoint returns a collection's items up to 100 per request with a CMS read scope, so a script can page through everything and write it out as Markdown or into a headless CMS.

Then the third column of the migration table, the part nobody quotes:

  • Forms. Webflow processes your submissions. Off Webflow, something else has to, and its routing needs testing end to end.
  • Redirects. Every old URL gets a one-to-one redirect, decided before launch. If the new site keeps the old paths, most of this disappears.
  • Interactions. Visual animations have to be rewritten in code. GSAP being free helps; it doesn't do the work.
  • Tracking. Every event gets rebuilt and re-verified on the new site. If the old site was tracking nothing useful, as ours was, this is the moment to fix it properly.
  • The build pipeline. Merging is the deploy, a check catches drift, and someone owns it. Our WordPress-to-static write-up covers the pipeline in detail.

There's also a middle path we haven't used ourselves. Webflow Cloud deploys Next.js, Astro and Vite apps either alongside a Webflow site, served under a path on its domain, or as independent apps. If marketing genuinely needs the Designer and you need one custom tool, that keeps both under one roof. Treat it as documented, not tested by us.

What an Astro site needs that Webflow gave you

Leaving a hosted builder means taking back jobs the vendor was doing. Name an owner for each before you start:

  • An editing story. Markdown in the repo for developer-run sites. A headless CMS for anything an editor touches weekly. The coworking operator chose an AI assistant in a terminal for occasional edits, which is a real option in 2026 for an owner who's comfortable describing a change and checking the result. It is not a visual editor, and a marketing team that rebuilds pages weekly will feel the difference.
  • Guardrails against drift. A static site's default failure is duplicated markup slowly going out of sync. In our prototype for the coworking operator, the location pages are generated from one data list, so they can't drift apart structurally. On our own site, the header and footer are written into every page from one partial, and the deploy fails if any page has drifted.
  • Hosting and deploys. Cheap and quiet, but someone has to own the account, the domain and the deploy key.
  • Forms, search and anything dynamic. Each is a small service you choose rather than a checkbox you tick.

Done properly, the result is very calm to run: no server to patch, no plugin to update, no login page to attack. Done carelessly, it's a site only one person can change.

Our defaults

Since a comparison that won't pick is a cop-out, here's how we'd settle Webflow vs Astro:

  • A marketing team builds and rearranges pages without a developer → stay on Webflow. Fix structured data, canonicals and tracking in place, through custom code or Webflow's MCP server.
  • Layouts rarely change, a developer or agency makes every change, and speed and control matter → Astro.
  • The site has to integrate deeply with a booking, membership or other platform → Astro or another custom build, with the platform keeping payments and logins.
  • You need both a visual editor and custom code → Webflow for the marketing pages and a custom app under one path, or an Astro front end with a headless CMS for the editors.
  • You want to leave because of search → don't. No platform is inherently better for SEO. Fix the content, structure and page head where you are first.

Related comparisons: Netlify vs Cloudflare Pages vs WP Engine: where to host · Squarespace vs WordPress for service businesses: stay or go?

If you're on Webflow and wondering whether it's still earning its keep, tell us who edits the site and how often. Sometimes the answer is a redesign or rebuild; often it's a week of fixes on the platform you already have.

Common questions

Is Astro better than Webflow for a marketing website?

Only for some teams. Webflow is better when non-technical people change layouts themselves, because its visual editor is the product and the hosting, security and publishing come with it. Astro is better when layout changes are rare and a developer makes them, because you get very fast static pages, full control of the page head and nothing to log into or patch. The deciding question is who adds a new section or page type, and how often. Answer that honestly and the choice usually makes itself. If the honest answer is nobody, you are paying for a visual editor you do not use.

Can I export my Webflow site to Astro?

Partly. As of November 2026, Webflow's code export is a feature of its paid Workspace plans, and it gives you the markup Webflow generated, not an Astro project. CMS content comes out separately, for example through Webflow's Data API, which lists a collection's items up to 100 per request. In practice you rebuild the layouts as Astro components, move the CMS content into Markdown or a headless CMS, then rebuild forms, redirects, interactions and tracking. The export is a reference, not a migration. Budget for the rebuild, and keep the old URLs where you can so fewer redirects are needed.

Who can edit an Astro website after launch?

Whoever your editing setup lets in, and that has to be decided before the build. Astro is a framework, not a CMS. Content can live in Markdown files in a repository, which suits developers, or in a headless CMS such as WordPress, Sanity, Storyblok, Decap, Keystatic or TinaCMS, which suits editors. Some owners now make changes by asking an AI assistant working in the repository. That works for occasional edits, but it is not a substitute for a visual editor if a marketing team rebuilds pages weekly. Pick the editing setup with the people who will use it, not for them.

Is Webflow or Astro better for SEO?

Neither inherently. Both can serve fast, well-structured pages, and rankings come from content, structure, internal linking and technical correctness. The practical difference is control of the page head. On Webflow, canonicals and structured data go in per-page custom code, and the Data API has no canonical field, so automating them means an OAuth app or Webflow's MCP server. On Astro they are files the build generates. If your site needs a lot of head-level work, that difference matters. If it does not, choose on who edits the site and leave SEO out of the platform decision.

Why does my Webflow site send Google Analytics hits to its own domain?

Webflow's GA4 integration can send hits first-party. On a Webflow site we audited, every analytics event was posted to a path on the site's own host rather than to Google's collection domain, with only an empty transport check going to Google. It matters mainly for testing: a test harness that blocks Google's hostnames to avoid polluting your data will still let real hits through. Block by request pattern instead, and answer every outgoing analytics request locally while you test. Also check which hostname each hit reports, because staging can send to the same property.

Can I use Webflow and Astro together?

Yes, in a few ways. As of November 2026, Webflow Cloud deploys Next.js, Astro and Vite apps either alongside a Webflow site, served under a path on its domain, or as independent apps. That lets marketing keep its pages in Webflow while a developer builds a custom tool under one path. The other route is a static Astro front end with a headless CMS. We have not shipped on Webflow Cloud ourselves, so treat that option as documented, not tested by us. Either way, decide which system owns each page before anyone builds.

Is Astro still open source now that it has joined Cloudflare?

Yes. The Astro Technology Company joined Cloudflare on 16 January 2026, and Astro's own announcement says the framework stays open source and MIT-licensed, keeps an open governance model, and continues to support a wide set of deployment targets rather than only Cloudflare's. For a marketing site that means the usual hosts for static files still work, and the code you own today stays yours. Watch the project's changelog, as you would with any dependency, but the licence is not the risk. The bigger risk on any static build is having no editing setup.