Translating a whole site automatically is straightforward; translating it well requires protecting structure and terminology.
1. Never translate raw HTML
Extract text nodes first, translate them, then rebuild the document. Translating markup directly corrupts attributes and can alter URLs.
2. Split long content into safe segments
Send paragraphs and list items individually rather than whole pages. Segment-level requests retry cleanly, which matters when a provider rate-limits you mid-job.
3. Protect what must not change
- Product and brand names.
- Placeholders, shortcodes and template variables.
- URLs, email addresses and code samples.
- Units, model numbers and version strings.
Replace each with a unique token before sending, then restore it afterwards.
4. Translate short fields first
Titles, summaries and navigation labels deliver most of the perceived localisation with the least risk. Body copy can follow once the structural pieces are verified.
5. Review a sample, not everything
Check a fixed sample per section: one homepage block, one long article, one edge case. If the sample is clean, spot check the rest rather than reading all of it.
6. Keep the source of truth
Store translations alongside the original text keyed by segment, so a changed paragraph can be re-translated without redoing the page.
Comments (0)
Log in to join the discussion
Log InNo comments yet