Blog · Myth vs. Data

Moving a Hand-Coded Site Into WordPress: What Five Converters Carried Over

We ran five HTML-to-WordPress converters on one made-up four-page site, scored on a rubric fixed before the first run. Here's what came through.

Maybe your membership site was hand-coded years ago, by you or by a friend who knew HTML. It works. But every small change, a new price on the services page or a new photo on the about page, means opening the code or waiting on the one person who can. WordPress would let you make those edits yourself, in an editor.

The thought that may stop you is simple: moving would mean building the whole site again.

We wanted to know how true that is today.

What this means for your site

You don’t have to start from a blank page, and you shouldn’t expect a perfect copy either. In this test of five converters, every tool brought some of a hand-coded site into WordPress and none brought all of it. Plan to check every page in the editor and to redo the look, the forms, the video and anything a script did. If you’d rather not do that part by hand, that’s our next piece: we’ll move a real site with the MemberPress.ai assistant. That one uses our own tool, and we’ll mark it that way. It’s coming soon.

Why we looked

WordPress Agent Benchmarks asks a plain question every month: when someone asks an AI agent for a website and doesn’t name a platform, does the agent choose WordPress? Its October 2026 recommendations say: “More and more people ask AI agents to build their websites, and unless they name WordPress, agents rarely reach for it.” (We read that line from the page’s data file on 2026-10-28.) The recommendations page is worth your time if you’re weighing where your site should live.

Part of that question is what happens to a site that already exists. If moving it means starting over, WordPress loses before anyone opens the editor. So we tested the part we could measure: what the tools that turn HTML into WordPress blocks actually carry over.

What we tested

We built one made-up site, Quillfern Cycle Works, a bike repair shop in a made-up town. It has four pages: home, about, services and contact. It’s plain HTML with one stylesheet and one shared script, and it has the things converters tend to drop, on purpose: a shared menu and footer, a photo gallery with captions, a price table, a booking form, a YouTube video and a small script that says whether the shop is open.

Five converters ran on it on 2026-10-28, each on its own fresh, throwaway copy of WordPress, never a live site. We used one fixed version of each tool, named with the scores below. The scoring rubric was written and saved before the first run: ten criteria, four pages, and every cell scores 0, 1 or 2. The sample site, the rubric, the scripts, every screenshot and every score are public in the test’s repo.

What came through, in plain facts

The look and the links. Only Static Site Importer carried the site’s font and colours into WordPress, and only it pointed the menu and page links at the new WordPress pages. It was also the only tool given the whole site, stylesheet included. The four page tools got each page’s HTML without the stylesheet, so their pages wear the default WordPress theme, and their links still point at the original pages.

Opening the pages in the editor. WordPress builds a page out of blocks, one for each heading, paragraph, picture or form. An invalid block is one the editor can’t read: it shows a warning and offers to recover it instead of letting you edit it. A Custom HTML block holds the original code as it was, so the page can look right, but you change it by editing code, not in the editor.

  • rawHandler: no invalid blocks. Every page landed as four Custom HTML blocks, holding the original code as it was.
  • Static Site Importer: the home page’s hero banner showed “Block contains unexpected or invalid content” in the editor.
  • Blocks Engine: no invalid blocks. On the home and contact pages, one block showed as a block this WordPress doesn’t have. The contact form came through with no form fields: the labels became paragraphs, and the only control left was the submit button.
  • Block Runner: no invalid blocks and no recovery notices on any of the four pages. The contact form kept all six controls, inside a Custom HTML block.
  • HTML to Blocks Converter: the four gallery images on the about page and two headings on the services page were reported invalid. The contact form is gone: no form fields are left, only a button.

The video and the shared script. No tool’s YouTube player kept its 16:9 shape at both screen widths. Blocks Engine’s was 16:9 on a wide screen and not on a phone. The site’s shared script, the one that opens the menu on a phone, ran on no converted page from any of the five tools.

The scores, and how we kept it fair

The scores, tool by tool

Static Site Importer (†) was given a different input, so don’t compare its number with the other four.

Tool, publisher and version testedAll ten criteriaN/EWithout the two capped criteria
rawHandler, WordPress (Gutenberg project), bundled in WordPress 7.1.425 / 72017 / 56
Static Site Importer †, Automattic, v1.33.17941 / 606not shown
Blocks Engine, Automattic, PHP Transformer v0.44.21735 / 66327 / 50
Block Runner, Human Made, 1.0.027 / 66319 / 50
HTML to Blocks Converter, Chris Huber, v0.7.231 / 72024 / 56

N/E is the number of cells with no evidence: cells the saved evidence couldn’t decide, left out of both the score and the maximum, which is why the maximums differ. † marks a different route and PHP version, explained under “How we kept it fair”.

The two capped criteria are the two checks the four page tools couldn’t fully pass whatever they did, because of the way we handed them each page: criterion 1 (structure and headings) and criterion 5 (links and nav). The last column leaves those two out. Static Site Importer has no number there, because the way we handed it the site didn’t hold it back on them.

Criteria 1 and 5 are worth 8 points each across the four pages, 16 in all, so each maximum drops by 16. rawHandler: 25 minus 4 and 4 is 17, out of 72 minus 16, which is 56. Blocks Engine: 35 minus 4 and 4 is 27, out of 66 minus 16, which is 50. Block Runner: 27 minus 4 and 4 is 19, out of 50. HTML to Blocks Converter: 31 minus 3 and 4 is 24, out of 56. Every cell is in results/SCORES.csv.

How we kept it fair
  • How we handed each tool the page. The four page tools got each page’s HTML, with relative links made absolute against the original copy, which is what you get when you copy from a live page. Nothing else was changed, and nothing was repaired by hand afterwards. Static Site Importer got the whole site as one ZIP file, the input it’s built for.
  • What that held back. Each page tool’s HTML became one WordPress page under the default theme. That theme prints the page title as a second main heading, which fails the “one main heading” check in criterion 1. And their links point at the original pages, which fails criterion 5. That’s why the table has a last column without those two.
  • The scoring notes. The rubric couldn’t foresee every case the runs turned up, so the scoring notes set out how each one is read, the same way for all five tools. They were written after the runs and before any score.
  • One tool ran a different way. Static Site Importer ran by its publisher’s own route in WordPress Playground, under PHP 8.5. The other four ran under PHP 8.3. We tried its command-line route first, and it failed twice in Playground. Both failed attempts are kept in the repo.
  • A second read. After the scores were saved, 40 cells were scored again from the evidence without looking at the first scores. None disagreed.

Before you move anything, open each page of your own site and list the forms, the video and the scripts it leans on. That list is what you will need to redo by hand, or hand to someone who can.

Sources

See it on your own site

Get your membership to its first paying member.

Connect your site and tell the assistant what you sell. It sets up your Join and Checkout pages and tells you what’s left before anyone can pay, like connecting your own Stripe or PayPal. Your password never leaves your site.

Connect your AI with Membership MCP→