Duplicate posts
WPML clones your content per locale. URLs become hostage to a plugin's schema. Migrate off it and your translations die.
paragraphs.co.uk
The translation backend for content-driven websites.
One content-addressable translation graph. Every CMS. Every framework. Edge-delivered.
Source unit · fingerprint A3F5C9E27K2P
<button data-pg="A3F5C9E27K2P">Add to cart</button>→ rendered as
Añadir al carrito at /es/Built by BuiltByGo. Used in production by
- TopReg
- TrustPadel
- FootballPrizes
- BuiltByGo
Works with WordPress · Next.js · Astro · Sanity · Nuxt · roll-your-own via API.
Translation is still broken in 2026.
Three pains every team running a multilingual site recognises.
Runtime ≠ source
Native framework i18n is a runtime — it doesn't know how to translate your CMS, your product catalogue, or your JSON-LD.
Brand drift
DeepL alone won't enforce your brand voice across 500 product names. ChatGPT alone is too expensive and inconsistent.
A translation graph that lives above your CMS.
Source content gets fingerprinted on save. The graph holds translations. Any frontend reads from it. Move CMS, the translations come with you.
01
Source
WP, Sanity, MDX, JSON
02
Fingerprint
blake3(unit_type || content)
03
Graph
source_units · translations · TM · glossary
04
Cascade
DeepL · Claude · TM lookup
05
Delivery
Edge Worker · build-time · hybrid
Edit one paragraph in a 50-block post — only that fingerprint re-translates. Switch from WordPress to Sanity next year — the same graph still serves your site.
Three architectural decisions you can't bolt on later.
Content-addressable.
Edit one paragraph; only that fingerprint re-translates. No locale-locked duplicate posts. Trivial CMS migration.
Multi-CMS by design.
WordPress plugin, Next.js adapter, Astro integration, Sanity studio plugin. One graph feeds all of them.
AISEO native.
Per-locale llms.txt. JSON-LD translated structurally. LLM Context API. Visible to ChatGPT and Google alike.
Your framework's i18n stays the runtime. We're the backend.
Paragraphs doesn't replace next-intl, Astro i18n, or @nuxtjs/i18n. It feeds them. That's the whole point.
What native i18n is good at
- · Locale routing and middleware
- · Server-side rendering per locale
- · ICU MessageFormat at runtime
- · Formatters for date / number / currency
What it can't answer
- · Where do the translations come from?
- · Who reviews them? Who versions them?
- · How does my CMS content get into a `messages/` file?
- · How do I keep brand glossary enforced across 12 locales?
Built for translation at scale.
Every feature serves the graph. Not a marketing buzzword among them.
Translation graph
Content-addressable storage. Same paragraph, same fingerprint, translated once.
Learn more
Model cascade
DeepL for short units, Claude for nuance and long-form. Cost-optimised by default.
Learn more
Translation memory
Embedded source units re-use prior translations. Cuts repeat cost to zero.
Learn more
Glossary
DNT, forced, suggested. Brand consistency enforced as model placeholders.
Learn more
Branches
Stage seasonal campaigns. Diff. Merge atomically. Works like git for translations.
Learn more
In-context editor
Click any translated element on your live site. Edit it. TM and glossary inline.
Learn more
Edge delivery
Cloudflare Workers substitute translations before the response reaches the browser.
Learn more
Flagship
AISEO
Per-locale llms.txt, translated JSON-LD, LLM Context API. Built for ChatGPT and Google alike.
Learn more
Plug it in wherever you already are.
WordPress
PHP plugin. Gutenberg blocks fingerprinted on save. WooCommerce-aware.
Learn more
Next.js
Postbuild script writes to messages/. next-intl stays your runtime.
Learn more
Astro
Astro integration auto-syncs translations to your content collections.
Learn more
WooCommerce
Product names, attributes, variations, transactional emails — all translated.
Learn more
Sanity
Studio plugin streams translations as documents update.
Learn more
Roll your own
OpenAPI spec, TypeScript SDK, REST + webhooks. Build whatever you need.
Learn more
Translations LLM crawlers actually understand.
Most translation plugins still optimise only for Google. We don't. Every locale has its own llms.txt, its own translated JSON-LD, its own LLM Context API endpoint.
-
Per-locale llms.txt
Auto-generated from approved translations. Refreshed daily. Surfaces your translated content to GPTBot, ClaudeBot, PerplexityBot.
-
Structurally translated JSON-LD
Schema.org properties translated as data, not strings. Product names, FAQ answers, BlogPosting bodies — all locale-correct.
-
LLM Context API
GET /api/llm-context?locale=es-ES returns a clean, machine-readable site summary. Cite-ready for LLM answers.
A serious tool, built on serious infrastructure.
Frontend
- WordPress · Next.js · Astro · Nuxt · Sanity
Edge
- Cloudflare Workers
- KV cache · R2 backups
Backend
- Supabase Postgres + pgvector
- Stripe Billing · WorkOS SSO
- · Cloudflare Workers, 300+ cities
- · Supabase Postgres 17 + pgvector
- · Stripe Billing + metered usage
- · EU + US data residency at signup
- · SOC 2 Type I in progress (Year 1)
- · GDPR Article 28 DPA published
- · WorkOS SSO for Enterprise
- · R2 backups with 90-day Object Lock
Predictable, word-based, no surprises.
Free
Evaluate the product on a real site.
£0 /mo
billed annually
- Words / month 100,000
- Locales 3
- Projects 1
- Users 2
- WordPress plugin + Next.js adapter
- Edge delivery in 300+ cities
- AISEO surfaces (llms.txt, JSON-LD)
- Glossary + Translation Memory
- Community support
Starter
For small sites going live.
£16 /mo
billed annually
- Words / month 500,000
- Locales 8
- Projects 3
- Users 5
- Everything in Free
- All headless adapters (Astro, Nuxt, Sanity)
- Branches
- Email support
- Top-ups available
Growth
Where most teams sit.
£65 /mo
billed annually
- Words / month 2,000,000
- Locales 20
- Projects 10
- Users 15
- Everything in Starter
- In-context editor
- Per-project model overrides
- Priority email support
- Audit log (24 months)
Start free in 30 seconds.
100,000 words / month on the free tier. No card. No call needed.