Why Loqua
Most translation apps still run on Google. We never do.
The AI features on every other Shopify translation app are a metered layer over a generic engine. The pricing forces the fallback. Here is the math, and why it quietly damages your brand.
The hidden pricing math
Take Transcy, one of the most widely installed Shopify translation apps, on around 64,000 stores. Its AI translation is metered by the token, at roughly 1.2 tokens per word. The $99 Cross-Border plan includes 1,000 tokens a month. The top Global plan is $599 for 5,000.
A thousand tokens covers on the order of 800 words at that rate, roughly a few dozen product descriptions rather than a storefront. By our own estimate, a modest D2C store with 75 SKUs across five European locales needs somewhere between 5,500 and 10,000 translated strings for an accurate first pass (products, collections, menus, theme strings, pages, metafields and SEO tags, multiplied across locales). A store running that many locales sits on Growth or Scale, not Starter.
Per Transcy's own documentation, content beyond your AI-token balance is handled by Google Translate at no extra cost. That is the pricing model, not a starter limitation you grow out of. The cheaper your plan, the more of your store a generic engine writes.
Transcy pricing as published on transcy.io, June 2026. Plans and token counts change over time; check their current pricing before comparing.
Why generic engines drift
English is full of meaning-rich words your brand uses creatively. A tank in your activewear catalogue is a sleeveless top to a shopper, a military vehicle to a generic engine. A trainer is a running shoe in one store and a fitness coach in another. Generic translation reaches for the dictionary default every time, in every locale. Product titles drift, category pages start to read foreign, and shoppers bounce before they trust you enough to buy.
The fix is not more AI tokens. It is removing the metering, and giving every translation the context it needs: your glossary, your voice profile, prior translations of similar items, and the product category itself.
That is what Loqua does. Every string passes through Claude with the full brand context. There is no generic fallback, because there is no quota to exhaust.
What an all-LLM store actually gives you
When you set up Loqua you give us three things: a brand voice profile (tone, formality, register), a glossary of terms that must translate a specific way, and a set of target locales.
Every string in your store, from product titles down to delivery descriptions, then runs through Claude with that context attached. Each translation gets a confidence score. High-confidence translations publish automatically. Anything the model is unsure about, or anything that conflicts with your glossary, lands in a review queue for a human to approve. Once approved, it enters your translation memory and never has to be generated again.
Identical and near-identical source content is never re-translated. Edit a product title and only the changed strings refresh, across every active locale. Accuracy builds as your catalogue settles, and your cost per month trends down even as the store grows.
See it side by side with your current app.
The clearest way to judge translation quality is to compare it. Walk through how Loqua works, or put it head to head with Transcy.