TranslatePress with a real backend.
TranslatePress had the right architectural instinct — a string table, not duplicate posts. We took that further: external graph, edge delivery, model cascade, AISEO. Same UX you liked. Better foundations.
What TranslatePress got right.
TranslatePress avoided the duplicate-posts trap. Its string table approach gives a much cleaner mental model — and a useful in-context visual editor. For a single WP site with stable content and no headless ambitions, it's a reasonable choice.
The architecture caps the ceiling.
-
Single-CMS by design
The string table lives inside WordPress. Move to Sanity, Next.js, or a mobile app, and your translations don't come with you.
-
No edge layer
Translation happens in PHP on every request. Caches are fragile; TTFB suffers. Paragraphs substitutes in a Cloudflare Worker, in <40ms p99.
-
No AISEO surfaces
No per-locale llms.txt, no LLM Context API, no structurally translated JSON-LD. Fine in 2022. Not fine in 2026.
Feature matrix.
| Feature | Paragraphs | TranslatePress |
|---|---|---|
| Storage model | External graph, fingerprinted | String table inside WP DB |
| Headless support | Yes | No |
| Edge delivery | Cloudflare Worker, <40ms p99 | PHP-rendered, no edge layer |
| Auto-translate engine | DeepL + Claude cascade | DeepL or Google |
| Translation memory | Yes | No |
| Glossary enforcement | Yes | No |
| Branches / staging | Yes | No |
| JSON-LD translation | Structural | Strings only |
| AISEO surfaces | Per-locale llms.txt + Context API | None |
| In-context editing | Yes | Yes |
Like TranslatePress's mental model?
You'll like ours — same idea, headless-first execution. Bring your translations over in a single import.