Shopify own LLM for store translations: check credits at the provider

You already pay a model. French titles are still English. A second AI invoice inside Shopify feels like paying twice. Shopify own LLM for store translations is Settings → AI connection in OO - Ultimate AI Translations: OpenAI-compatible base URL, API key, model. The app bills $0 for that AI. When a job fails with an AI error, you look at *your* provider — key, model name, remaining credits. Not Shopify Billing.

Listing 5 Nov 2026: Free $0 (connect your own AI, unlimited languages/markets/translations, language switcher; no backups, no scheduled jobs, App AI not included). Professional $10 + usage. $20 Plus plan + usage. App AI ~$0.01/unit, billed at the translation cost. 0 reviews. Launched 14 October 2026. Do not invent stars.

Related: connecting an OpenAI-compatible API is the three fields. Usage costs is the App AI meter. Yesterday: daily jobs on Plus. This piece is the own-model path on Free, and what an AI error actually is.

Locales do not fill copy. Your model does.

Languages / Markets publish locales. They do not write product titles, theme strings, or metaobjects. Shopify’s own machine translation is a different product. It is not a Settings screen with a base URL you choose, and it is not a lock that survives the next job.

Connect AI: the app needs a provider for Translate actions and automatic jobs.

Two paths:

1. Your API. OpenAI-compatible base URL, key, model. Save. A small test translation. Listing: Connect your own AI — on Free, and on paid as “Or use your own AI”.

2. App AI. Professional/Plus only. Default if you leave the Home setup on it. Usage ~$0.01/unit. Pass-through: billed at the translation cost.

OpenAI-compatible means the endpoint speaks that protocol. Listing “Works with”: A.I., ChatGPT, LLMs. That is not a warranty that every model on the internet works. Use the URL your provider documents, the key you created there, the model string that account expects. Do not invent a model id.

Brand voice and glossary ride on the request so the brand name does not respell itself each run. Editor saves stay locked; jobs will not overwrite them. Jobs skip handles, URLs, IDs, SKUs, size codes, JSON, empty source, locked fields.

What the docs mean by “AI error”

Exact line: *If a job fails with an AI error, check your key, model name, and remaining credits with your provider.*

Three checks, that order.

Key. Wrong key, revoked key, key from another project. The app stores what you pasted. Shopify Billing will not show this.

Model name. Typo, a model your account cannot call, a retired string. The job fails. The catalogue stays source language. Same three fields, correct string, retry.

Credits at the provider. Own LLM means own balance. $0 in the app is not unlimited at OpenAI or whoever you wired. Empty is empty. Translate stops. That is not a Jobs bug. Top up, or on Professional/Plus switch to App AI (~$0.01/unit) until the wallet recovers.

Do not mix that up with:

• Skipped fields: empty, locked, technical. FAQ: not an AI error.

• A schedule that never fired: timezone, disabled, no language/content type, AI not connected. FAQ.

• Free with no clock: daily/weekly/monthly are paid. Own LLM on Free runs Translate actions, not overnight jobs.

Own LLM versus App AI

Own LLM fits when the key already exists, you want to stay on Free, or you want the app’s usage meter at $0 on $10/$20. Unlimited translations on Free means unlimited *in the app*, not unlimited at the provider.

App AI fits when you do not want to babysit a key, or provider credits are awkward. Professional: monthly schedule, glossary, 1+1 backup. Plus ($20): daily, 5+5 backups. Your own API remains available on both.

Not: Free + App AI. Not: expecting jobs to translate SKUs. Not: reading an AI error as “the lock broke” — locks protect edits; they do not empty a provider wallet.

The language switcher is on Free (embed or theme block). Backups are not. Do not claim them.

Worked example: the wallet that hit zero overnight

Shop “Ridge & Pack”. EN source, FR going live. An OpenAI-compatible key already runs internal tools. Free: publish FR. Settings → AI connection: base URL, key, model. Save. Translate Missing on one collection, titles + SEO.

Result: SKUs skipped (correct). Two titles failed, AI error. Provider dashboard: credits at 0 since yesterday. No Shopify charge. Top up, same three fields, retry, green. Glossary already pinned the brand; that run did not respell it.

A typo in the model string would have been the same column: AI error, not skipped. FAQ skipped is a different job-details bucket.

Then: lock the two titles the copywriter rewrote. A later Translate Outdated leaves them. Credits stay a provider problem.

Do not upgrade to Professional because “the LLM broke”. It was empty. Upgrade for monthly jobs or glossary features Free does not include.

Get a translation on the page

Install OO - Ultimate AI Translations.

1. Publish the locale you want filled.

2. Settings → AI connection: OpenAI-compatible URL, key, model. Save.

3. Smoke test: a handful of Missing fields, not the whole catalogue.

4. Job details: translated / skipped / failed. Failed + AI error → key, model, credits *at the provider*.

5. Switcher: embed or block on the live theme.

Free if you have the key. Professional $10 for monthly + glossary + one backup, with or without App AI. Plus $20 for daily. Usage only while App AI runs.

Wrap

Shopify own LLM for store translations is Free (and paid, if you keep your API): three fields, $0 AI inside the app, unlimited languages on the listing. An AI error is key, model name, or leftover credits at your provider. App AI on $10/$20 is ~$0.01/unit pass-through, optional.

Wire the model in OO - Ultimate AI Translations and run ten titles before you throw the whole FR catalogue at an empty wallet.