Yes. Register any post type and SleekAI can include it in the prompt context. Episodes, videos, courses, talks, ebooks, even bespoke types like "interview transcripts" all work the same way as standard posts. ACF and Meta Box fields on those CPTs flow in too, so you can recommend by speaker, by series, or by reading level.

If you allow it in the system prompt - many teams keep it scoped to their own archive to avoid sending traffic away. A common middle ground is to let the bot reference external sources by name when an internal piece doesn't exist yet, but never link out. That keeps the user oriented toward your archive while acknowledging the broader landscape.

Content flows in live from the database - no separate index to maintain. A piece you publish at 9am can be recommended at 9:01. If you cache the prompt context for performance you'll see the new piece on the next cache rebuild. Most teams set the cache window to match their publishing rhythm.

Yes. Multibot lets you scope a bot to a category, tag, or template, and display conditions handle the routing. The marketing-blog bot can have a different prompt and a different content scope than the engineering-blog bot, even on the same WordPress install.

Configure access by role or post status - paid content can be hidden, teased with a soft paywall, or fully exposed depending on the visitor's tier. For membership sites this is the same logic that gates the rest of the experience, applied to the chatbot's grounding scope.

Built-in logs let you filter by topic, query, and outcome. The most useful filter is "queries that didn't return a confident match" - those are the topics the archive doesn't cover well, which is exactly the editorial backlog the recommender is supposed to surface.

Limit the bot to two or three suggestions per reply, with reasoning, and let the user ask for more. Avoid the temptation to push a paid course or a newsletter signup in every reply. Trust comes from relevance; the conversion comes after the third or fourth genuinely useful suggestion.

Yes, via webhook or the JS API. A common pattern is to let the bot offer the newsletter only after two or three exchanges, when engagement is real. That converts better than the popup that interrupts the first article and respects the reader's attention.