Not needed. SleekAI generates each reply directly in the active language by instructing the model accordingly, so there is no double-translation pass through Weglot. Replies feel native because they are native, not run through a translation memory after the fact. That also means chat content never adds to your Weglot translation backlog, which keeps the multilingual setup simple.

By default the model answers in the active Weglot locale, which is what most stores want; a French visitor on the French page expects French replies even if they typed in English by mistake. You can configure the bot to mirror the visitor's input language instead through the prompt, which suits multilingual support desks where staff speak many languages and customers code-switch freely.

Yes. Set per-language welcome and placeholder text directly in SleekAI. A French visitor sees a French opener; a German visitor sees a German one. The same applies to preset suggested questions, which means "Bonjour, livrez-vous en France?" is a clickable preset for French visitors instead of an English version that has been auto-translated and feels off.

Posts come from WordPress, and Weglot's layer translates the front-end view at request time. SleekAI reads the WordPress source content and asks the model to reply in the active language. The bot's grounding is the original post; the localization is the model's output language. This keeps the prompt small and avoids the cost of indexing every translated variant when one canonical version plus a locale instruction does the same job.

No. Chat replies are not translated by Weglot, so they never count against your Weglot plan. Weglot's pricing is keyed to translated front-end content, and SleekAI generates chat replies natively in the locale through the LLM instead of routing them through the translation layer. For high-volume support bots that produce thousands of replies a month, that distinction matters; otherwise the Weglot bill would scale linearly with chat usage.

Any OpenAI, Anthropic, Google, or OpenRouter model that supports the language you need. The major European languages, plus Japanese, Chinese, Arabic, and Hebrew, are well-served by current frontier models. Less common languages may need careful prompt testing; SleekAI's conversation logs are how you spot quality issues per locale and decide whether to switch models for specific languages or stick with one provider across all.

Yes. The widget respects the locale direction set by Weglot, so Arabic and Hebrew sessions render right-to-left automatically. Layout, message alignment, and input direction all flip. If you customize the widget styling, double-check that any custom CSS includes RTL-friendly equivalents so the widget stays readable in both directions without manual overrides per locale.

Conversation logs record the active language alongside every chat, so you can filter logs by locale and spot problem areas. A common pattern is to review the bottom-quartile language by user satisfaction or by escalation rate, then either tighten the prompt for that language or switch to a model with better coverage. Without the per-language tag in the log, this kind of targeted tuning is impossible.