No. SleekAI reads WordPress content directly through the icl_translations join and asks the model to reply in the active language. Translation memory is what WPML uses to suggest translations for editors, not a runtime layer for chat replies. The bot's grounding is the published translated post; the model handles language fluency natively without needing a memory pass.

The chat widget supports RTL automatically when WPML sets the locale direction. Arabic, Hebrew, Persian, and Urdu pages flip the widget layout, message alignment, and input direction. Custom widget CSS should include RTL equivalents so styling stays consistent. WPML handles the locale direction signal; SleekAI consumes it without a separate config switch.

Yes. Display conditions filter by WPML language code, so you can run a different bot per language. The common pattern is a single bot for most languages with a specialized bot for a high-value locale, like a Japanese-only bot with a tighter prompt because the company has a focused presence there. Mixing global and locale-specific bots in the same setup is supported.

Yes. UI strings, including the chat welcome message, placeholder text, and preset suggested questions, can be translated through WPML String Translation. The bot picks up the translated string for the active language at render time. That means a French visitor sees French labels, a German visitor sees German labels, and there is no parallel translation system to maintain inside SleekAI.

Yes. WPML's WooCommerce Multilingual extension translates products, attributes, and meta. SleekAI reads the translated product for the active language so pre-sale chats reference the right product description, price, and stock state. A French shopper on a French product page gets French answers grounded in the French product, not in the English source product. Currency switching is handled by WPML; the bot consumes the active currency context.

OpenAI, Anthropic, Google, and OpenRouter, with your own key. Pick the model that performs best for your most-used languages. Frontier OpenAI and Anthropic models are excellent across major European languages, Japanese, and Chinese. Less common languages may benefit from a model with stronger multilingual training; SleekAI's conversation logs let you compare quality per language and switch providers if needed.

WPML returns the source post when no translation exists, and SleekAI uses that fallback. The model is still instructed to reply in the active language, so it translates the source content into the visitor's language at generation time. The result is usually fine for incidental content gaps, but for important pages the right fix is to publish the translation in WPML rather than relying on the model's translation pass.

WPML and Yoast or RankMath handle the multilingual sitemap; SleekAI is a chat surface, not a sitemap generator. What matters for the bot is that translated posts are accessible through the WPML API at runtime, which they are. If your SEO sitemap exposes translated URLs, the bot reads the same translations through icl_translations and answers consistently across both surfaces. The two systems coexist without conflict.