Yes. Movement caliber, case material, case size, lug-to-lug, lug width, water resistance, power reserve, and any other field you store in WooCommerce attributes or custom meta become part of the chatbot's context. The bot quotes those values precisely rather than describing watches in the abstract. For shops that maintain detailed spec sheets, this is one of the few categories where chatbot accuracy directly maps to conversion: shoppers who get the right lug-to-lug answer book the order; shoppers who get a guess do not.

If your products use WooCommerce variations for dial color, strap material, or bracelet versus strap, the bot reads stock per variation at request time. A customer asking about the green dial sees current stock for the green-dial variation, not the aggregate stock of the SKU. For limited editions or numbered pieces, you can flag the variation accordingly and the bot will say "3 of 50 still available" rather than a generic in-stock status.

Drop your warranty terms and service-interval page into the chatbot's context (or into OpenAI Files for longer documents) and the bot quotes the policy verbatim. Service intervals for common Sellita, ETA, Miyota, and Seiko movements can live in a structured table the bot reads. The result is an answer like "the SW200-1 is recommended for service every 5 to 7 years; our workshop partner charges 280 EUR" rather than a vague "please contact us about service."

Yes, within the bounds of what you publish. If your product copy includes a regulated-accuracy claim ("regulated in 5 positions to +/- 6 sec/day"), the bot quotes that. For chronometer-certified pieces, COSC standards and certificate references can live in product meta and the bot surfaces them. Avoid asking the bot to extrapolate beyond what you have published, since accuracy claims are something customers will hold the shop to in writing.

Yes. If a customer is browsing the German version of the site, the bot loads the German product copy and policy pages and replies in German. The system prompt can also be language-specific, so the tone matches local norms. Watch shops with EU, UK, US, and JP audiences typically run one bot per language with shared structured context (movement caliber tables, water resistance ratings) and language-specific marketing copy.

If you publish a fit guide (case size, lug-to-lug, wrist circumference ranges), the bot can read it and surface candidates that match the customer's wrist size. The same logic works for strap suggestions by lug width. The bot stays factual: it surfaces options drawn from the catalog rather than pushing one SKU. Logs from these conversations are useful for refining the fit guide itself, since the bot's misses are often the gaps in the published copy.

In your WordPress database. Each message is logged with model name, token usage, and page URL. Nothing routes through Sleek infrastructure: conversations go from the browser to your WordPress site to your model provider on your API key. For shops with EU customers, this keeps the data-processor map short and the data-residency story clear. Logs can be retained or purged on the schedule your privacy policy describes.

Multibot lets you run separate bots with different system prompts, display rules, and even different model providers. An entry-level field-watch bot can be approachable and educational; a haute-horlogerie bot can be more technical and reserved. Display conditions cover category, tag, user role, logged-in state, and URL pattern, so each bot only appears on the pages where it belongs. Customers see one widget at a time, with the tone that matches the piece they are looking at.