How to redesign a WordPress website with AI

Where AI actually helps

Three places, reliably. First, the audit: crawling every URL, flagging thin pages and clustering what ranks takes an afternoon with AI instead of a week. Second, first-draft copy: block-level drafts on a defined voice, which a human then cuts by a third. Third, QA: AI sweeps for broken links, missing alt text, orphan pages and schema errors faster than any manual pass.

Where AI hurts if you let it

The conversion strategy. AI will happily generate a homepage that reads like every other homepage generated this year. The distinction the brand draws itself, the proof with names and numbers attached, the offer architecture: that is human work. Same for the redirect map. An unmapped URL is a ranking you paid for and then deleted, and no model feels that loss for you.

The working split

1. AI audits, you decide. Crawl, cluster, flag. Then a human picks what lives, dies or merges.
2. System first, then generation. Lock tokens, type and blocks before any AI touches a page, or every page drifts.
3. AI drafts inside the system. Copy generated block by block against the voice rules, then edited down.
4. Humans own the redirect map. Every URL mapped one-to-one before launch. This step is non-negotiable.
5. AI sweeps, humans sign off. Links, alt text, schema and speed checked by machine; the launch call stays human.

What this looks like on a real project

On a recent build, the catalogue alone was 1,546 items taken from Figma to live on WordPress. The inventory, clustering and QA at that scale are AI work by necessity. The template decisions, the proof placement and the redirect plan were human throughout. That split is the whole article in one sentence.