Exporting a Webflow site you don't own the Workspace for: what a 62-post CMS looks like from the outside

Webflow's own CMS export is a CSV per collection, from the Designer, and it needs access to the project. If you only have the live site (a client who left, a freelancer who disappeared, a site you inherited), the published HTML is what is left. It is static, and it says more than it seems to.

We ran an export against marketing.flowbase.co, a public Webflow site, on 31 October 2026, and read the files that came out. These are the numbers and the traps. We did not import the result into a WordPress install for this note: what follows is what the files contain, not what a given importer does with them.

The inventory

• 62 posts (33 in /blog, 29 in /article, two collections that both read like a blog) and 21 pages
• 577 images, about 94 MB, all on Webflow's CDN
• About 28,000 words in the posts
• Hundreds of CMS items that are not posts: 229 in /clone, 24 in /tools, 17 in /booster, 8 in /support

1. The CMS leaves marks in the HTML

Every Webflow page has data-wf-site on the <html> tag. A page that comes from a CMS collection also has data-wf-collection and data-wf-item-slug. That is how you tell a post from a normal page without guessing from the URL. The sitemap then gives the list: a collection is a folder with the same shape, /<collection>/<slug>.

2. Which collection is the blog?

Nothing in the HTML says “this collection is the blog”. Here two collections have blog names, so both are exported as posts. On other sites the name is anything. Our rule is the names people use (blog, posts, articles, news, journal…), and if none matches, the largest collection. It is a heuristic. A shop's product collection or a directory of 229 clones is not a blog, and a site whose blog is called /stuff next to a bigger collection of something else will pick the wrong one. That is why the scan lists the other collections by name and count instead of hiding them.

3. The date is a design decision

Webflow sites have no standard place for the publish date: each designer puts it in a text element. We look for structured data, then a <time> tag, then an element whose class says date. In this site, 33 posts had one. The other 29 (the help articles) show no date at all, so the only date left is the sitemap's lastmod, which is when the item was last published. Those posts get the date and a warning in the inventory, so nobody believes it is the original date.

4. The body is the biggest rich-text field

A CMS item's article lives in a .w-richtext element, but a page can have several (summaries, author bios, captions). The biggest by text length is the article. Headings, lists, tables, links and images survive; -p-500-style resized copies of images are mapped back to the original file name. 28 pages and posts had a custom code embed that cannot move; each one carries a warning in the inventory.

5. Pages repeat the same newsletter block

The first run put the same newsletter section on 16 pages, with its three decorative letter images each time. The form itself is dropped (forms can't be exported), but the heading and the images stayed. Short sections that contain a form are now left out of pages, with a warning, instead of being copied 16 times into the media library.

What the visitor's view cannot give you

• Categories, authors and tags: in the CMS they are references to other collections, and the page only shows their names as text.
• Forms, e-commerce and memberships.
• Custom code embeds, interactions and the layout itself. Content moves; the design does not.
• Draft and unpublished items, and anything behind a password.

A checklist before you copy anything

• Find out which collections are posts and which are not; do not trust the folder name alone.
• Check dates: if the page shows none, the sitemap's lastmod is the last publish, not the first.
• Save the images, not just the links: they live on Webflow's CDN and go away with the project.
• Map every old URL (/blog/<slug>) to the new one before you move the domain.
• Rebuild forms and embeds by hand, and write them down first.

Decamp does the reading for you and writes WordPress, Ghost, Markdown and CSV files plus the redirects. See exporting a Webflow site, Webflow to WordPress or Webflow to Ghost.

Paste a Webflow address to count its posts, pages and images for free.