ChatGPT Deep Research
An agentic research capability that autonomously browses the web, synthesizing hundreds of sources into comprehensive cited reports. Now powered by the GPT-5.5 model family.
What it is
Agentic research capability where ChatGPT autonomously browses the web to find, analyze, and synthesize information from hundreds of sources into a structured, cited report. Launched in February 2025 for Pro users and expanded to all tiers by April 2025. Now powered by the GPT-5.5 model family (the latest models), delivering stronger reasoning and synthesis quality than the original o3 version. The agent can interpret text, images, and PDFs during its research. A real time progress sidebar shows which sources are being read and what reasoning steps are being taken. As of February 2026, Deep Research can connect to apps (formerly called connectors) and restrict searches to trusted sites. Connected apps include Google Drive, SharePoint, and authenticated industry data sources like FactSet, PitchBook, and Scholar Gateway. A fullscreen document viewer lets you read and navigate completed reports in an expanded view. Legacy Deep Research mode was retired on March 26, 2026. Reports can be downloaded in Markdown, Word, and PDF formats. Benchmark performance: 26.6% on Humanity's Last Exam (vs o1 at 9.1%) and 72.57% on GAIA, demonstrating strong multi step reasoning and research capability.
Who it is for
Researchers needing comprehensive literature reviews with hundreds of cited sources
Business professionals conducting market research, competitive analysis, or due diligence
Students writing research papers who need thorough, multi source investigation with proper citations
Anyone who needs a comprehensive multi source investigation and is willing to wait 5 to 30 minutes for thoroughness
Pros and cons
Pros
- +Produces comprehensive reports that would take hours to compile manually, complete with structured sections and inline citations from hundreds of sources
- +Inline citations from hundreds of sources let you verify every claim and dig deeper into any specific finding
- +Real time progress tracking shows exactly what the agent is doing, so you can monitor the investigation and redirect if needed
- +Available on the Free tier with limited access, making it accessible for occasional research needs
- +Now powered by the GPT-5.5 model family, delivering stronger reasoning and synthesis quality compared to earlier models
- +Connected apps (February 2026) allow the agent to access Google Drive, SharePoint, and authenticated industry data sources like FactSet and PitchBook during research
- +Fullscreen document viewer provides an improved reading experience for lengthy research outputs
- +Can process uploaded files (PDFs, spreadsheets, documents), so the agent builds on your existing knowledge rather than starting from zero
Cons
- −Slow by design: each research session takes 5 to 30 minutes, making it unsuitable for time sensitive questions that need immediate answers
- −Limited queries on Free and Plus (10/month) tiers; heavy researchers will hit limits quickly and need Pro $100 (50/month) or Pro $200 (unlimited)
- −Pro $100 at $100/mo or Pro $200 at $200/mo is expensive if Deep Research is your primary use case
- −Cannot access paywalled content; the agent can only read freely accessible web pages
- −No API access for Deep Research specifically; developers cannot integrate this capability into their own applications
Pricing
- Limited Deep Research queries
- Structured reports with citations
- Real time progress tracking
- 10 Deep Research sessions/month
- Powered by GPT-5.5 models
- Connected apps (Google Drive, SharePoint, etc.)
- File upload for context
- 50 Deep Research sessions/month
- 5x the Plus usage quota
- GPT-5.5 Pro model access
- Unlimited Deep Research sessions
- 20x the Plus usage quota
- GPT-5.5 Pro model access
Prices change; check the official pricing →