Skip to main content
When you ask a question that requires external information, the assistant launches a research subagent to find answers. Each subagent runs independently, searches or analyzes data, and returns a sourced report directly in your chat thread. There are two types of research subagents: web research and connected account insights.

Web research

Web research searches the internet for information relevant to your question. It fetches pages, extracts key content, and produces a report with source links. How to trigger it: Ask about trends, competitors, industry topics, or anything that requires up-to-date external information.
Research the top 3 trends in sustainable fashion this week and show me a summary with sources.
What are my competitors doing on Instagram in the fitness space? Find recent examples.
You can focus research on a specific domain or website by mentioning it in your prompt.
Search techcrunch.com for recent articles about AI-powered marketing tools.

Connected account insights

Connected account insights analyze your linked social media accounts — engagement rates, top-performing posts, content patterns, and audience trends. How to trigger it: Ask about your own account performance on any connected platform (X, Instagram, LinkedIn, YouTube, and others).
Analyze my Instagram account performance and give me insights on what's working.
What are my top-performing posts on X this month?
You must have a platform connected before the assistant can analyze it. See Connect platforms to link your accounts.

Reading research cards

When a research subagent runs, a research card appears in your chat thread.

Card header

  • Icon: Globe icon for web research, chart icon for connected account analysis.
  • Status badge: Shows the current state of the research.
  • Elapsed time: How long the subagent has been running.
  • Token usage circle: Shows how much of your context the research consumed.

Status badges

BadgeMeaning
ResearchingThe subagent is actively searching or analyzing.
CompletedResearch finished successfully. Expand to read the report.
FailedSomething went wrong. Retry or adjust your prompt.
InterruptedThe subagent stopped unexpectedly. Try again.

Collapsed and expanded views

  • Collapsed: Shows the status badge, a brief preview of what the subagent is doing, and elapsed time.
  • Expanded: Click to expand and see the full research transcript — each search query, page fetch, and intermediate finding as it happened.

Report and citations

Once research completes, the card footer shows:
  • Research report: The full findings, collapsible for easy scanning.
  • Citations: Source links for web research. Click any citation to visit the original page.
  • Copy button: Copy the report text to your clipboard.

Aggregate reports

When your question triggers multiple research subagents — for example, comparing performance across platforms — the assistant can synthesize their individual reports into a single aggregate report. The aggregate report includes:
  • Cross-platform overview combining all findings.
  • Platform comparison (engagement, content types, audience).
  • Top-performing content across sources.
  • Strategic recommendations based on the combined data.
Citations from all individual reports are deduplicated and listed together.
Compare my X and Instagram performance side by side and give me a unified strategy.

Tips

  • Be specific about your research goal. “Research sustainable fashion trends in the US market for Gen Z” gives better results than “research fashion.”
  • Use domain focus for targeted web research. Mention a specific site or domain to narrow results.
  • Ask for cross-platform analysis when you have multiple accounts connected to get aggregate insights.
  • Research subagents consume basic tokens. Each subagent launch costs 1 basic token, plus additional tokens as the research grows. See Understand token usage for details.