Deep Research transforms ChatGPT into a comprehensive research tool that conducts multi-step web searches, synthesizes findings from dozens of sources, and produces structured reports with citations. It goes far beyond a single search query, following research trails across multiple domains to build a complete picture of any topic.
With the ability to pause mid-search, refine your queries, connect to internal data through Connectors, and export polished PDF reports, Deep Research delivers the kind of thorough analysis that previously required a dedicated research team.
ChatGPT Deep Research autonomously searches the web across dozens of sources, follows citation trails, cross-references findings, and produces structured reports. Key features include the ability to pause mid-search and redirect the research direction, Connectors integration for accessing internal company data alongside public information, and PDF export for sharing polished reports with stakeholders. The legacy mode was removed on March 26, 2026, and all users now access the current experience with enhanced interactive controls and output quality.
A research workflow designed for depth, accuracy, and user control at every stage.
When you submit a research query, Deep Research does not return a single set of search results. Instead, it plans a research strategy, executes multiple search queries across different angles, reads and analyzes the content of relevant pages, identifies gaps in the information gathered, and conducts follow-up searches to fill those gaps. This iterative process mirrors how an experienced human researcher works, producing results that are comprehensive rather than surface-level.
The system evaluates source quality, cross-references claims across multiple publications, identifies contradictions in the literature, and synthesizes everything into a coherent narrative. The final output includes structured sections with clear headings, inline citations linking back to source material, and a summary of key findings. This depth of analysis is particularly valuable for market research, competitive analysis, academic literature reviews, and due diligence investigations where incomplete information can lead to poor decisions.
One of Deep Research's most powerful capabilities is the ability to pause the research process at any point and intervene. As the system reports its progress, you can review intermediate findings and realize the research is heading in the wrong direction, or that a particular sub-topic deserves deeper exploration. Pause the search, provide new instructions, narrow the scope, add specific questions you want answered, or redirect the entire research focus.
This interactive control distinguishes Deep Research from automated report generators that produce fixed outputs. A market analyst might start with a broad industry overview, pause when an unexpected competitor appears in the results, and redirect the research to focus specifically on that competitor's strategy. A journalist might narrow the search to a specific time period after seeing initial results span too broadly. The pause-and-refine workflow respects the reality that research questions evolve as you learn more about a topic, following the approach described in OpenAI's Deep Research documentation.
For organizations that need to combine public research with internal knowledge, Connectors bridge the gap. Configure Deep Research to access your company's internal databases, knowledge management systems, document repositories, CRM data, or enterprise tools. When conducting research, the system pulls relevant information from both public web sources and your connected internal data, producing reports that reflect your organization's specific context and proprietary intelligence.
Connectors are particularly valuable for enterprise teams conducting competitive intelligence, internal audits, or strategic planning where public data alone is insufficient. A sales team can research a prospect and simultaneously pull internal CRM notes and past interaction history. A compliance team can cross-reference regulatory changes with internal policy documents to identify gaps. All connector data remains within your organization's security perimeter, with the same encryption and access controls that protect Enterprise plan data.
Deep Research generates structured reports that export as polished PDF documents ready for professional distribution. The exported PDF includes formatted sections with clear headings, data tables, citation lists with links to source material, key findings summaries, and optional appendices. The output is designed for stakeholder presentations, board reports, academic submissions, and archival purposes where a permanent, shareable document format is required.
The PDF export saves hours of manual formatting that would otherwise follow a research phase. Instead of copying findings from ChatGPT into a document editor and restructuring them, Deep Research delivers the final format directly. Teams use this capability to produce weekly market briefings, quarterly competitive analyses, and ad-hoc research reports for leadership requests with minimal post-processing.
The original Deep Research experience has been fully replaced with the current enhanced version.
OpenAI removed Deep Research legacy mode on March 26, 2026. The legacy mode used an older search and synthesis pipeline that lacked pause-and-refine capabilities, Connectors integration, and PDF export. All users who previously relied on legacy mode were automatically migrated to the current experience. No action was required, and existing saved research outputs remained accessible in chat history. The current Deep Research pipeline delivers faster search execution, higher-quality synthesis, better source evaluation, and the interactive controls that make the feature genuinely useful for professional research workflows.
Complete breakdown of what Deep Research can do and how each capability serves different research needs.
| Capability | Description | Available On |
|---|---|---|
| Multi-Step Web Search | Autonomous multi-query research across dozens of sources | Plus, Pro, Business, Enterprise |
| Pause and Refine | Interrupt research to redirect, narrow, or expand scope | Plus, Pro, Business, Enterprise |
| Source Cross-Reference | Verify claims across multiple independent publications | Plus, Pro, Business, Enterprise |
| Connectors Integration | Access internal databases and enterprise knowledge bases | Business, Enterprise |
| PDF Export | Generate formatted reports with citations and tables | Plus, Pro, Business, Enterprise |
| Citation Tracking | Inline source references with links to original material | Plus, Pro, Business, Enterprise |
| Structured Output | Organized sections, key findings, executive summaries | Plus, Pro, Business, Enterprise |
| Research History | Save and revisit previous research sessions | Plus, Pro, Business, Enterprise |
Stop spending hours on manual research. Deep Research delivers comprehensive, cited reports in minutes. Available on Plus and higher plans.
Get StartedDetailed answers about Deep Research capabilities, access, and the legacy mode transition.
Deep Research is an advanced research feature in ChatGPT that conducts autonomous multi-step web searches, synthesizes findings from dozens of sources, cross-references information for accuracy, and produces structured reports with inline citations. It goes beyond simple search by following research trails, identifying information gaps, and conducting follow-up searches to build comprehensive analyses on any topic.
Yes. Deep Research provides interactive control throughout the research process. You can pause at any point, review intermediate findings, and redirect the search. Narrow the scope, add new questions, change the research focus, or request deeper exploration of a specific sub-topic. This ensures the final report matches your exact needs rather than following a rigid, predetermined search path.
Connectors allow Deep Research to access your organization's internal data sources alongside public web information. Business and Enterprise plans can configure Connectors to internal databases, knowledge bases, CRM systems, and document repositories. Research reports then combine proprietary internal intelligence with public findings, creating comprehensive analyses that reflect your organization's specific context and data.
Yes. Deep Research generates structured reports that export as polished PDF documents with formatted sections, data tables, citation lists, key findings summaries, and optional appendices. The PDF output is designed for stakeholder sharing, board presentations, academic submissions, and archival purposes. No manual reformatting is required after export.
OpenAI removed Deep Research legacy mode on March 26, 2026. The legacy pipeline has been fully replaced by the current experience with pause-and-refine controls, Connectors integration, and PDF export. All users were migrated automatically with no action required. Existing saved research outputs from legacy mode remain accessible in chat history.
Features that complement Deep Research for complete analytical workflows.
ChatGPT