It can sit alongside the existing search form or fully replace it. Most teams keep both visible during a transition and watch the logs to see which path visitors prefer. Once the AI search proves out, hiding the legacy form is one display-condition rule away.

Yes. Any registered post type, taxonomy, or postmeta key can be included in the prompt context. ACF, Meta Box, and Pods fields all work because they live in postmeta. You decide which fields are worth the token cost - usually the ones your editors actually fill in.

The widget is lazy-loaded so the search bar itself adds almost nothing to the initial page weight. The database query that fetches grounding context only runs when a visitor sends a message, and the request goes straight to your model provider rather than back through WordPress repeatedly.

Yes. SleekAI is BYO API key, so you pay your provider directly at their rates. That keeps the cost transparent and lets you switch providers if pricing changes. For a typical site, search-bar usage costs are low because the prompts are small and answers are short.

Yes. Restrict by post type, taxonomy, status, role, or any custom WP_Query argument. You can exclude draft pages, password-protected content, or anything tagged "internal". Display conditions add a second layer so different sections of the site search different content scopes.

SleekAI works with WPML, Polylang, and TranslatePress. The active locale is detected on the request, so the bot reads the matching translation and replies in the same language. You can also assign a per-language system prompt via Multibot to tune tone and formality.

WooCommerce variations and stock status live in postmeta, so they flow into the prompt the same way as any other field. The bot can answer "is the medium in stock" or "which colour ships fastest" if you include the relevant meta keys. Keep the field list lean to control token cost.

Yes. Logs include the query, the grounding context, the reply, and any handoff signal. Filter by failed grounding or short replies to find the queries the archive can't answer. That list is the most useful editorial backlog you'll find on a content site.