How I, a strawberry farmer, moved my farm's website from WordPress to EmDash on Cloudflare with Claude Code

On the night of October 21, the rental server I was using shut down every site it hosted as a precaution against a PHP vulnerability. A smaller, separate site could no longer be fixed either.
This is the record of how, within a week, I moved my strawberry farm's website from WordPress to EmDash, which runs on Cloudflare Workers. I did the work together with Claude Code. I had never written a program: my first site was a WordPress theme I bought, with CSS I asked ChatGPT for. Every number below comes from my work records.
Practice first, then my own site
On October 22, I first tried the move on a smaller, separate site. It worked, so I decided the same day to move my farm's site too. I think it was right not to start with the farm site itself. The mistakes from the smaller site became the checklist for the farm site.
What the old WordPress site looked like
- Many plugins
- About 800 lines of custom PHP
- 48 theme-specific components inside the page content
- 65 HTML blocks
My goal was simple: make the site easy to edit from an admin screen and with AI.
I feel that many strawberry-picking farm websites are built once and then left alone for years. Even farms that are active on social media often have a homepage that has not changed in years. I think that is because editing is a hassle, or because every change costs money. What mattered most to me was a site I can manage myself, and that AI can help me edit from now on.
The timeline
I split the work into stages: prepare, build locally, soft launch, switch over, and watch for two to four weeks.
- Oct 23: soft launch, hidden from search engines
- Oct 24 (a rainy Saturday): switched over around noon
- Oct 28: stopped the old WordPress site (the files were kept)
Protecting search traffic
Many visitors find a strawberry-picking farm through search, so not losing rankings was the top priority.
- The 63 pages that appeared in search kept exactly the same URLs, character for character.
- Old URLs for images, categories, tags and so on were sent to their new locations with 301 redirects: 1,464 in total.
- Before the switch, I recorded the title, meta description, canonical, h1 and structured data of every page. After the switch, I compared them by script. There were zero differences.
- I did not change page titles until mid-November, so that if rankings moved, I could tell whether it was the migration or my content edits. I edited the body text every day.
- I prepared a rollback plan before switching: recreating two DNS records would bring the old site back.

I also made the site faster with automatic image resizing, font subsetting and caching. And I keep a rule: no fake reviews or star ratings.
How I worked with Claude Code
The most useful decision was to split the roles before starting.
- Me (the human): DNS, passkeys, Search Console, and the decision to publish.
- Claude: the migration, redirects, structured data, and before-and-after comparisons.
Other practices:
- I turned the failures from the smaller site into a table, then into a checklist for each stage.
- Every change to the live site was written as a dated script and kept in the history.
- Claude made two or three design mockups, and I chose one.
- I asked a second AI (Codex) to review the diffs and to list what was missing compared with standard practice.
Where I stumbled
- Some redirects were lost in the migration and those pages returned 404s. I added them later.
- The structured data still contained an old price.
- Right after the switch, my local DNS cache showed the old site, and I almost thought the switch had failed. I checked the response headers to confirm the new site was live.
- The first deploy failed with an unclear error. The second one succeeded.
- The built-in image resizing was not working in production.
- Portrait photos from an iPhone came out with the wrong orientation after conversion.
- Automatic emails from a no-reply@ address landed in spam.
Results
Mobile performance (Lighthouse, median of three runs):
- Home page: 55 → 95
- Strawberry-picking page: 62 → 95
- Article page: 67 → 94
The same page can vary between 68 and 96 from run to run, so I treat these as rough guides.
In Search Console, I compared 18 days before and 18 days after the switch. Average position and CTR stayed the same. Impressions and clicks rose 37%, but that includes the season and advance bookings. So I don't claim the migration increased traffic. My conclusion is that the migration did not cause a drop.
Advice if you want to do the same
- Don't start with your live site. Trying a smaller site first turns mistakes into a checklist.
- Decide how to roll back before you switch.
- Keep URLs and titles unchanged until things settle.
- Split roles between human and AI. Keep publishing decisions and credentials with the human; give the repetitive work and the checking to the AI.
Who WordPress suits better
If you want AI to write and edit your articles, I think EmDash is the better fit. But it is not for everyone. WordPress suits you better if you:
- want to get things done with plugins and themes
- would rather not touch code or AI tools
- value the large amount of information and the many agencies you can hire
- would rather not depend on Cloudflare alone
EmDash has weak points too: it is hard to move away from Cloudflare, it is still new with little information, and there are few plugins.
I chose EmDash because I plan to keep improving the site together with AI.
If you choose WordPress, you will need a rental server (web hosting).
Closing
From the night the server went down to the day I stopped the old WordPress site took exactly one week. Now I have a site I can edit every day with an admin screen and AI. I hope this record helps other farmers and small shops with an aging website.
The full story, in Japanese: いちご農家の私が、Claude Code で農園のサイトを WordPress から EmDash(Cloudflare)へ移した記録