No. Hotjar's own tracker still runs and captures heatmaps, recordings, and survey responses. SleekAI sits beside it and reads the metadata you publish into WordPress about those artifacts, like which heatmap is live where, what the survey asked, and what the latest research note concluded.

By matching the current URL or page ID against a custom post type or option that lists active studies. Most teams maintain this inventory anyway. SleekAI's data-source wizard reads it and feeds the matching rows into the system message at request time, so the answer reflects the latest published state.

Only if those transcripts are in WordPress. The recommended pattern is to keep one-paragraph research summaries linked to recording IDs, not the full transcripts. That keeps the prompt small, protects respondent privacy, and gives the bot enough context to point a stakeholder at the right clip.

Not on every request. Sync a small daily or hourly summary of question, response count, and top themes into postmeta and let the bot reference that snapshot. This is faster, friendlier to Hotjar's rate limits, and easy to debug when a count looks off, because the value lives in a single WordPress field.

Yes. Use multibot with display conditions on user role and URL pattern. Internal admins see a detailed bot that quotes research notes and contributor emails. Public visitors see a high-level bot that explains how feedback is collected without exposing individual findings.

Yes. SleekAI's multibot mode lets you run a Hotjar research bot, a Mixpanel tracking-plan bot, and a public support bot in parallel. Each has its own system prompt, data sources, and display conditions, so the right bot loads automatically based on who is viewing and where.

Hotjar's own consent and PII rules still apply to recordings. SleekAI only reads what you explicitly map into the prompt, so transcripts, identifiable session data, and raw responses do not leak through unless you expose them yourself. For most teams, exposing summaries and study metadata is enough and keeps the privacy surface small.

Any supported by SleekAI's BYO-key providers: OpenAI, Anthropic, Google, OpenRouter. Research Q&A is mostly retrieval over a small inventory of studies and notes, so a fast cheap model is fine for most questions. Reserve a stronger model for nuanced asks like comparing two findings or summarising a long research roundup.