Skip to main content
Research Chat is an AI-powered panel built into every study page. Ask questions about your conversation data in natural language and get answers grounded in actual user interviews, complete with clickable source references.

​
Accessing Research Chat

Open any study from your dashboard. The Research Chat panel is always visible on the right side of the page — no navigation required. When you first open it, you see a prompt input with rotating placeholder suggestions and quick prompt buttons.

​
Asking questions

Type any question about your study data: 
What are users saying about pricing?
What features do users request most?
How do users feel about onboarding?
Pillow generates an answer based on your actual conversations. Each response includes source references — click any source to jump to the original conversation transcript.
Specific questions like “What frustrates users about checkout?” yield richer answers than broad ones.

​
Quick prompts

Pre-built prompts help you get started:
  • Summarize key findings — Overview of what your conversations reveal
  • Executive summary — High-level narrative for stakeholders
  • Most requested features — What users want added or improved
  • Conversations we should read — The most valuable interviews your team should review
  • Most frequent issues — Recurring problems across conversations
Click any quick prompt to generate an instant analysis.

​
Managing chat conversations

Research Chat uses Chrome-style tabs to manage multiple conversation threads:
  • New tab: Click the + button to start a fresh conversation
  • Switch tabs: Click any tab to switch between threads
  • Close tabs: Click the x on a tab to close it
  • History: Use the dropdown to browse and reopen past conversations
Each tab maintains its own conversation history, so you can explore different angles of your study simultaneously.

​
Adding context

You can attach specific context to your chat messages to focus the analysis:
  • Conversation context: From the conversation list, add a specific conversation as context to your next question
  • Q&A context: From the Q&A tab, add a specific question cluster as context
Context appears as a pill above the input field showing what’s attached. Click the x on the pill to remove it before sending.

​
Visual evidence in answers

Research Chat answers can include QA distribution charts inline — bar charts and pie charts drawn from your Q&A analysis. These appear automatically when the answer references a specific question cluster, giving you visual evidence alongside the text explanation.

​
Filtering by metadata

Narrow Research Chat answers to a specific audience segment. Click the filter icon in the chat input to select metadata values:
  • Locale: Filter by user language
  • Country: Filter by geographic location
  • Device: Filter by desktop, mobile, or tablet
  • Custom fields: Filter by any metadata you passed via URL parameters
When a filter is active, all answers are scoped to conversations matching your criteria. Remove filters to return to the full dataset.

​
Sharing chat output

Use Share to copy a link to what you want teammates to review:
  • Share entire conversation: Click Share in the top-right corner of the chat panel
  • Share one answer: Click Share on a specific assistant message
When creating a link, choose:
  • Visibility: Organization-only or Public
  • Expiration: Never, 7 days, or 30 days
Organization-only links require sign-in to your organization. Public links can be opened by anyone with the URL.

​
Q&A analysis tab

The Q&A tab provides automated question-and-answer clustering across all your conversations.

​
How it works

Once your study reaches 100 conversations, Q&A analysis generates automatically. It groups similar questions your agent asked and shows how users responded. Each question cluster includes:
  • Question theme: The grouped question
  • Ask count: How often this question was asked
  • Answer buckets: How users typically responded, with percentage breakdowns

​
Regenerating

When new conversations come in after the last analysis, a banner shows how many new conversations are available. Click Regenerate to update the analysis with the latest data.
Q&A analysis requires at least 100 conversations to identify meaningful patterns. While waiting, you can use the chat tab to ask questions about any conversations you’ve collected so far.

​
Small sample banner

When your study has fewer than 10 conversations with replies, a banner reminds you to treat answers as early signals rather than established trends. The banner can be dismissed per conversation tab.

​
Tips for effective use

  1. Start with quick prompts to get a high-level view, then drill down with specific questions
  2. Use multiple tabs to explore different angles — one for product feedback, another for UX issues
  3. Follow up on answers to dig deeper into a specific finding
  4. Add context when you want the analysis to focus on a particular conversation or question cluster
  5. Share the executive summary with stakeholders who need a quick overview

​
Next steps

Review conversations

Read individual interviews in detail.

Study analysis concept

Understand how Q&A clustering and Research Chat work.