.svg){kind=logo}
CoverManager makes the booking software behind more than 16,000 restaurants across 25 countries, around a hundred of them Michelin-starred. Its content (blog, press room, guides) lived in WordPress, spread across six languages with wildly uneven volumes: Spanish had over 300 articles, English had one, and English was the language meant to lead the new site. French never made the move: it fell out of scope at kick-off.
The redesign was run by Digidop, who handled its design and Webflow development. That left the content, and the job was enormous: redoing the work by hand, language by language and article by article, meant several months of effort and a serious risk of breaking the site's search rankings along the way. Vydera came in to support that part: migrating, producing the missing versions and cleaning up the SEO.
The starting point: a lopsided multilingual site, an SEO to protect
The company wanted to leave WordPress for Webflow to gain editorial independence and brand consistency. On the content side, the goal fit in one sentence: move everything across without losing the search rankings already earned, and use the move to clean the content up rather than copy the mess over. Two contradictory demands, because the safest way not to break SEO is to touch nothing at all.
Six languages, one real body of content
Articles per language in WordPress (grey), then on Webflow after migration (green). The new site was meant to have English as its primary language. It had 1 article. French was not carried over.
First real decision: do not migrate everything. Of nearly 300 articles reviewed, 239 deserved to stay. The other 54, outdated or duplicated, were dropped. The migration started with an editorial cull.
A migration starts with a cull
Of 293 blog articles reviewed, 54 were outdated or duplicates. They were dropped rather than moved.
Why the manual approach did not hold
Done by hand, each difficulty made the other three worse.
- Volume: 239 articles, plus the press room, the guides and the site pages.
- Languages: each piece had to exist in five consistent versions, which meant producing the missing ones, not just moving what already existed.
- SEO: titles, meta descriptions, heading hierarchy, internal links, URL slugs, all things a sloppy migration breaks in silence.
- Webflow: the platform imposes a precise structure, collections, fields and templates, one language at a time.
Copy-paste, outsource the translation, re-import article by article: at this scale the conventional approach does not hold. One formatting mistake, multiplied by 239 articles and 5 languages, means a whole site to redo page by page.
The answer: a migration driven by Claude Code
Rather than handling pages one at a time, we drove the whole content migration with Claude Code: AI agents that read the source content, transform it and publish it straight through Webflow's API (Data API v2), without going through the editor.
The real work is producing the languages. We took the content from its most reliable source, Spanish, then localised it into the four other languages, keeping exactly the same structure from one version to the next. Where WordPress had a single article in English, the new site has 239, aligned with the Spanish version.
Our rule: do nothing irreversible "live" on the site. We prepare everything locally, we check it, then we replay a controlled import. If something looks wrong, we fix the script and run it again, without starting over and without creating duplicates.
Nothing irreversible, live on the site
Every piece of content follows the same route. If something looks wrong, we fix the script and replay it: an article already migrated is recognised, never recreated.
How we held the volume without giving up quality
Everything through the API, prepared locally
No manual copy-paste. Each article is converted offline, WordPress "block" content becoming clean Webflow rich text, then checked, then published via the API. The pipeline can be rerun safely: we keep a mapping between each source item and its Webflow identifier, so an article already migrated is recognised and skipped instead of being recreated. The import can be replayed as often as needed and always converges to the same state.
One source, five languages, one identical structure
The Spanish content is the reference; AI localises it into the four others. The rule that matters: the HTML structure stays identical from one language to the next, only the text changes. A guardrail checks that no tag, no embed, no table has moved during translation. In practice we create the article in the primary language, then roll out each language version, with its own slug and its own metas, without ever duplicating the article.
SEO rebuilt, piece by piece
This is what marketing teams actually care about:
- Internal links between articles were rewritten to point at the right page in the right language, using the site's real slugs, not URLs rebuilt by hand.
- Missing titles and meta descriptions were generated or completed.
- The heading hierarchy was put back in order, with no skipped level.
- Alt text was added to images in all 5 languages.
- Structured data, FAQ, breadcrumbs and related articles, was rolled out by template.
What protects everything already earned: the 301 redirects. Without them, every old WordPress address becomes a dead end and the accumulated ranking is lost. The redirect plan was drawn up by CoverManager's SEO freelancer, and we implemented it in Webflow. Content and redirects moved forward together.
Checked at every step
Every item sent is immediately read back via the API to confirm it matches what was expected. At the end, the number of published articles matches the migration log exactly: 239, with no orphan pages and no duplicates. Everything goes to staging first, and going live stays a decision.
The results
- 239 articles kept out of nearly 300 reviewed, localised into 5 languages, close to 1,200 versions.
- Press room (45 pieces of coverage), around twenty guides and about thirty pages migrated and localised in the same run.
- SEO put back in order across the board: titles, meta descriptions, structure, alt text, internal links, structured data, 301 redirects integrated.
- 0 publishing errors, every item read back automatically after writing. No duplicates, no ghost pages.
- 2 weeks of content migration, inside an 11-week redesign project, against several months for a manual rebuild.
The risk avoided
A manual migration at this scale almost always produces broken internal links, duplicate slugs, pages lost to Google, inconsistent language versions. Each of those costs rankings and weeks of rework. None of it happened: the site went live in a state verified end to end.
What it changes
CoverManager now has a multilingual, indexable content library, editable by its own teams in Webflow and ready to grow. SEO was not degraded by the move, it came out of it cleaner: that is the foundation a search strategy can now build on.
The invisible defects we fixed before going live
Migrating 239 articles in 5 languages mostly means facing a series of details that do not show at first glance and cost a lot once the site is live. A fast, unchecked migration lets them through. Three examples from this project.
- Page addresses held hostage by old versions. In Webflow, deleting a page in the working version does not remove it from the version already published. The old address stays "taken" and stops the new page from claiming the right URL, which is a problem when that URL carries the ranking. We cleaned the working version and the published version, in the right order, to free each address.
- Bulleted lists that vanished as soon as the article was reopened. Some nested lists displayed correctly on the public site but emptied themselves the moment an editor reopened the article in Webflow. The content held until the first edit. We rewrote those lists in the format the Webflow editor understands natively, across nearly 130 articles.
- A form sliding to the very bottom of the page. The contact form embedded in the articles had no anchor point in the code. With nowhere to attach, it was pushed down to the footer instead of appearing in the body. We put that anchor back on nearly 190 articles, in all 5 languages.
A systematic check detects these defects before going live. Otherwise you find them three months later, in Google's reports.
What to take away
- A successful migration starts with the cull. Dropping 54 outdated articles is worth more than moving them across neatly.
- Multilingual content is built, not translated. One reliable source language, a locked structure, and the four others generate without drifting.
- SEO does not survive on its own. Rebuilt internal links, completed metas and 301 redirects are one single job: miss one, and the ranking goes.
- What is not checked is not done. Reading every item back via the API after writing is what turns "we pushed it" into "it is live and correct".
How do you migrate a multilingual site from WordPress to Webflow without losing SEO?
By preparing the migration offline and replaying it through the API rather than doing it by hand. Internal links are rewritten to the real slugs of each language, titles and meta descriptions are kept or completed, a 301 redirect plan covers the old addresses, and every page is read back automatically after publishing to confirm no SEO element was lost.
Can AI localise a site into several languages reliably?
Yes, as long as you impose a structural constraint: the HTML layout stays identical from one language to the next, only the text is translated, and a guardrail checks that no tag has drifted. For CoverManager, Spanish was the source and AI produced the four other languages at constant structure.
How long does it take to migrate several hundred articles to Webflow?
For CoverManager, the content migration took about two weeks, inside an eleven-week redesign project. By hand, the same volume means several months. The gain comes from API automation and AI localisation, which absorb the volume without copy-pasting.
What does Claude Code bring to a Webflow migration?
Claude Code orchestrates AI agents that read, transform and publish content through Webflow's API, while handling localisation, SEO clean-up and verification. The whole thing is replayable: if something looks wrong, you fix it and run it again without creating duplicates.
How do you avoid broken links and duplicates in a Webflow migration?
By making the process idempotent, with each source item tied to its Webflow identifier and therefore never recreated twice, and by rewriting internal links from the slugs actually published. A written-versus-read-back reconciliation confirms, at the end, that the number of live pages matches the expected count exactly.
.jpeg)
